nimbus-eth2/tests/test_blockchain_dag.nim

1048 lines
36 KiB
Nim
Raw Normal View History

# beacon_chain
# Copyright (c) 2018-2022 Status Research & Development GmbH
# Licensed and distributed under either of
# * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
# * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
# at your option. This file may not be copied, modified, or distributed except according to those terms.
{.used.}
import
unittest2,
eth/keys, taskpools,
../beacon_chain/spec/datatypes/base,
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
../beacon_chain/spec/[beaconstate, forks, helpers, signatures, state_transition],
disentangle eth2 types from the ssz library (#2785) * reorganize ssz dependencies This PR continues the work in https://github.com/status-im/nimbus-eth2/pull/2646, https://github.com/status-im/nimbus-eth2/pull/2779 as well as past issues with serialization and type, to disentangle SSZ from eth2 and at the same time simplify imports and exports with a structured approach. The principal idea here is that when a library wants to introduce SSZ support, they do so via 3 files: * `ssz_codecs` which imports and reexports `codecs` - this covers the basic byte conversions and ensures no overloads get lost * `xxx_merkleization` imports and exports `merkleization` to specialize and get access to `hash_tree_root` and friends * `xxx_ssz_serialization` imports and exports `ssz_serialization` to specialize ssz for a specific library Those that need to interact with SSZ always import the `xxx_` versions of the modules and never `ssz` itself so as to keep imports simple and safe. This is similar to how the REST / JSON-RPC serializers are structured in that someone wanting to serialize spec types to REST-JSON will import `eth2_rest_serialization` and nothing else. * split up ssz into a core library that is independendent of eth2 types * rename `bytes_reader` to `codec` to highlight that it contains coding and decoding of bytes and native ssz types * remove tricky List init overload that causes compile issues * get rid of top-level ssz import * reenable merkleization tests * move some "standard" json serializers to spec * remove `ValidatorIndex` serialization for now * remove test_ssz_merkleization * add tests for over/underlong byte sequences * fix broken seq[byte] test - seq[byte] is not an SSZ type There are a few things this PR doesn't solve: * like #2646 this PR is weak on how to handle root and other dontSerialize fields that "sometimes" should be computed - the same problem appears in REST / JSON-RPC etc * Fix a build problem on macOS * Another way to fix the macOS builds Co-authored-by: Zahary Karadjov <zahary@gmail.com>
2021-08-18 18:57:58 +00:00
../beacon_chain/[beacon_chain_db],
../beacon_chain/consensus_object_pools/[
attestation_pool, blockchain_dag, block_quarantine, block_clearance],
./testutil, ./testdbutil, ./testblockutil
func `$`(x: BlockRef): string = shortLog(x)
2020-10-14 20:23:04 +00:00
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
const
nilPhase0Callback = OnPhase0BlockAdded(nil)
nilAltairCallback = OnAltairBlockAdded(nil)
nilBellatrixCallback = OnBellatrixBlockAdded(nil)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
proc pruneAtFinalization(dag: ChainDAGRef) =
if dag.needStateCachesAndForkChoicePruning():
dag.pruneStateCachesDAG()
suite "Block pool processing" & preset():
setup:
var
db = makeTestDB(SLOTS_PER_EPOCH)
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor, {})
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = Quarantine.init()
state = newClone(dag.headState)
cache = StateCache()
info = ForkedEpochInfo()
att0 = makeFullAttestations(state[], dag.tail.root, 0.Slot, cache)
b1 = addTestBlock(state[], cache, attestations = att0).phase0Data
b2 = addTestBlock(state[], cache).phase0Data
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
test "basic ops":
check:
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.getBlockRef(default Eth2Digest).isNone()
let
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
b0 = dag.getForkedBlock(dag.tail.root)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
bh = dag.getForkedBlock(dag.head.root)
bh2 = dag.getForkedBlock(dag.head.bid)
check:
b0.isSome()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
bh.isSome()
bh2.isSome()
dag.getBlockRef(dag.finalizedHead.blck.root).get() ==
dag.finalizedHead.blck
dag.getBlockRef(dag.head.root).get() == dag.head
test "Simple block add&get" & preset():
let
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
b1Get = dag.getForkedBlock(b1.root)
check:
b1Get.isSome()
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
b1Get.get().root == b1.root
b1Add[].root == b1Get.get().root
dag.heads.len == 1
dag.heads[0] == b1Add[]
let
b2Add = dag.addHeadBlock(verifier, b2, nilPhase0Callback)
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
b2Get = dag.getForkedBlock(b2.root)
sr = dag.findShufflingRef(b1Add[].bid, b1Add[].slot.epoch)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
er = dag.findEpochRef(b1Add[].bid, b1Add[].slot.epoch)
validators = getStateField(dag.headState, validators).lenu64()
check:
b2Get.isSome()
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
b2Get.get().root == b2.root
b2Add[].root == b2Get.get().root
dag.heads.len == 1
dag.heads[0] == b2Add[]
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.containsForkBlock(b2.root)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.parent(b2Add[].bid).get() == b1Add[].bid
# head not updated yet - getBlockIdAtSlot won't give those blocks
dag.getBlockIdAtSlot(b2Add[].slot).get() ==
BlockSlotId.init(dag.getBlockIdAtSlot(GENESIS_SLOT).get().bid, b2Add[].slot)
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
sr.isSome()
er.isSome()
# er reuses shuffling ref instance
er[].shufflingRef == sr[]
# Same epoch - same epochRef
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
er[] == dag.findEpochRef(b2Add[].bid, b2Add[].slot.epoch)[]
# Different epoch that was never processed
dag.findEpochRef(b1Add[].bid, b1Add[].slot.epoch + 1).isNone()
# ... but we know the shuffling already!
dag.findShufflingRef(b1Add[].bid, b1Add[].slot.epoch + 1).isSome()
dag.validatorKey(0'u64).isSome()
dag.validatorKey(validators - 1).isSome()
dag.validatorKey(validators).isNone()
# Skip one slot to get a gap
2020-05-19 15:46:29 +00:00
check:
process_slots(
Implement split preset/config support (#2710) * Implement split preset/config support This is the initial bulk refactor to introduce runtime config values in a number of places, somewhat replacing the existing mechanism of loading network metadata. It still needs more work, this is the initial refactor that introduces runtime configuration in some of the places that need it. The PR changes the way presets and constants work, to match the spec. In particular, a "preset" now refers to the compile-time configuration while a "cfg" or "RuntimeConfig" is the dynamic part. A single binary can support either mainnet or minimal, but not both. Support for other presets has been removed completely (can be readded, in case there's need). There's a number of outstanding tasks: * `SECONDS_PER_SLOT` still needs fixing * loading custom runtime configs needs redoing * checking constants against YAML file * yeerongpilly support `build/nimbus_beacon_node --network=yeerongpilly --discv5:no --log-level=DEBUG` * load fork epoch from config * fix fork digest sent in status * nicer error string for request failures * fix tools * one more * fixup * fixup * fixup * use "standard" network definition folder in local testnet Files are loaded from their standard locations, including genesis etc, to conform to the format used in the `eth2-networks` repo. * fix launch scripts, allow unknown config values * fix base config of rest test * cleanups * bundle mainnet config using common loader * fix spec links and names * only include supported preset in binary * drop yeerongpilly, add altair-devnet-0, support boot_enr.yaml
2021-07-12 13:01:38 +00:00
defaultRuntimeConfig, state[], getStateField(state[], slot) + 1, cache,
info, {}).isOk()
let
b4 = addTestBlock(state[], cache).phase0Data
b4Add = dag.addHeadBlock(verifier, b4, nilPhase0Callback)
check:
b4Add[].parent == b2Add[]
dag.updateHead(b4Add[], quarantine)
dag.pruneAtFinalization()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
check: # getBlockIdAtSlot operates on the head chain!
dag.getBlockIdAtSlot(b2Add[].slot).get() ==
BlockSlotId.init(b2Add[].bid, b2Add[].slot)
dag.parentOrSlot(dag.getBlockIdAtSlot(b2Add[].slot).get()).get() ==
BlockSlotId.init(b1Add[].bid, b2Add[].slot)
dag.parentOrSlot(dag.getBlockIdAtSlot(b2Add[].slot + 1).get()).get() ==
BlockSlotId.init(b2Add[].bid, b2Add[].slot)
var blocks: array[3, BlockId]
check:
dag.getBlockRange(Slot(0), 1, blocks.toOpenArray(0, 0)) == 0
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
blocks[0..<1] == [dag.tail]
dag.getBlockRange(Slot(0), 1, blocks.toOpenArray(0, 1)) == 0
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
blocks[0..<2] == [dag.tail, b1Add[].bid]
dag.getBlockRange(Slot(0), 2, blocks.toOpenArray(0, 1)) == 0
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
blocks[0..<2] == [dag.tail, b2Add[].bid]
dag.getBlockRange(Slot(0), 3, blocks.toOpenArray(0, 1)) == 1
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
blocks[1..<2] == [dag.tail] # block 3 is missing!
dag.getBlockRange(Slot(2), 2, blocks.toOpenArray(0, 1)) == 0
blocks[0..<2] == [b2Add[].bid, b4Add[].bid] # block 3 is missing!
2020-10-14 20:23:04 +00:00
# large skip step
dag.getBlockRange(Slot(0), uint64.high, blocks.toOpenArray(0, 2)) == 2
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
blocks[2..2] == [dag.tail]
2020-10-14 20:23:04 +00:00
# large skip step
dag.getBlockRange(Slot(2), uint64.high, blocks.toOpenArray(0, 1)) == 1
blocks[1..1] == [b2Add[].bid]
2020-10-14 20:23:04 +00:00
# empty length
dag.getBlockRange(Slot(2), 2, blocks.toOpenArray(0, -1)) == 0
# No blocks in sight
dag.getBlockRange(Slot(5), 1, blocks.toOpenArray(0, 1)) == 2
2020-10-14 20:23:04 +00:00
# No blocks in sight
dag.getBlockRange(Slot(uint64.high), 1, blocks.toOpenArray(0, 1)) == 2
# No blocks in sight either due to gaps
dag.getBlockRange(Slot(3), 2, blocks.toOpenArray(0, 1)) == 2
blocks[2..<2].len == 0
# A fork forces the clearance state to a point where it cannot be advanced
let
nextEpoch = dag.head.slot.epoch + 1
nextEpochSlot = nextEpoch.start_slot()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
parentBsi = dag.head.parent.atSlot(nextEpochSlot).toBlockSlotId().get()
stateCheckpoint = dag.stateCheckpoint(parentBsi)
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
parentBsi.bid == dag.head.parent.bid
parentBsi.slot == nextEpochSlot
# Pre-heated caches
dag.findShufflingRef(dag.head.parent.bid, dag.head.slot.epoch).isOk()
dag.findShufflingRef(dag.head.parent.bid, nextEpoch).isOk()
dag.getEpochRef(dag.head.parent, nextEpoch, true).isOk()
# Getting an EpochRef should not result in states being stored
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
db.getStateRoot(stateCheckpoint.bid.root, stateCheckpoint.slot).isErr()
# this is required for the test to work - it's not a "public"
# post-condition of getEpochRef
getStateField(dag.epochRefState, slot) == nextEpochSlot
assign(state[], dag.epochRefState)
let
bnext = addTestBlock(state[], cache).phase0Data
bnextAdd = dag.addHeadBlock(verifier, bnext, nilPhase0Callback)
check:
# Getting an EpochRef should not result in states being stored
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
db.getStateRoot(stateCheckpoint.bid.root, stateCheckpoint.slot).isOk()
test "Randao skip and non-skip":
process_slots(
dag.cfg, state[], getStateField(
state[], slot) + 1, cache, info, {}).expect("can advance 1")
let
proposer_index = get_beacon_proposer_index(
state[], cache, getStateField(state[], slot))
privKey = MockPrivKeys[proposer_index.get]
randao_reveal = get_epoch_signature(
getStateField(state[], fork),
getStateField(state[], genesis_validators_root),
getStateField(state[], slot).epoch, privKey).toValidatorSig()
var bad_randao_reveal = randao_reveal
bad_randao_reveal.blob[5] += 1
block: # bad randao + no skip = bad
let
tmpState = assignClone(state[])
message = makeBeaconBlock(
dag.cfg, tmpState[], proposer_index.get(),
bad_randao_reveal,
getStateField(tmpState[], eth1_data),
default(GraffitiBytes), @[], @[], BeaconBlockExits(),
default(SyncAggregate), default(ExecutionPayload),
noRollback, cache)
check: message.isErr
block: # bad randao + skip = bad
let
tmpState = assignClone(state[])
message = makeBeaconBlock(
dag.cfg, tmpState[], proposer_index.get(),
bad_randao_reveal,
getStateField(tmpState[], eth1_data),
default(GraffitiBytes), @[], @[], BeaconBlockExits(),
default(SyncAggregate), default(ExecutionPayload),
noRollback, cache, {skipRandaoVerification})
check: message.isErr
block: # infinity + no skip = bad
let
tmpState = assignClone(state[])
message = makeBeaconBlock(
dag.cfg, tmpState[], proposer_index.get(),
ValidatorSig.infinity(),
getStateField(tmpState[], eth1_data),
default(GraffitiBytes), @[], @[], BeaconBlockExits(),
default(SyncAggregate), default(ExecutionPayload),
noRollback, cache, {})
check: message.isErr
block: # Infinity + skip = ok!
let
tmpState = assignClone(state[])
message = makeBeaconBlock(
dag.cfg, tmpState[], proposer_index.get(),
ValidatorSig.infinity(),
getStateField(tmpState[], eth1_data),
default(GraffitiBytes), @[], @[], BeaconBlockExits(),
default(SyncAggregate), default(ExecutionPayload),
noRollback, cache, {skipRandaoVerification})
check: message.isOk
test "Adding the same block twice returns a Duplicate error" & preset():
let
b10 = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
b11 = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
check:
b11.error == BlockError.Duplicate
not b10[].isNil
test "updateHead updates head and headState" & preset():
let
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
dag.updateHead(b1Add[], quarantine)
dag.pruneAtFinalization()
check:
dag.head == b1Add[]
getStateField(dag.headState, slot) == b1Add[].slot
test "updateState sanity" & preset():
let
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
b2Add = dag.addHeadBlock(verifier, b2, nilPhase0Callback)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
bs1 = BlockSlotId.init(b1Add[].bid, b1.message.slot)
bs1_3 = BlockSlotId.init(b1Add[].bid, 3.Slot)
bs2_3 = BlockSlotId.init(b2Add[].bid, 3.Slot)
let tmpState = assignClone(dag.headState)
# move to specific block
var cache = StateCache()
check:
dag.updateState(tmpState[], bs1, false, cache)
tmpState[].latest_block_root == b1Add[].root
getStateField(tmpState[], slot) == bs1.slot
# Skip slots
check:
dag.updateState(tmpState[], bs1_3, false, cache) # skip slots
tmpState[].latest_block_root == b1Add[].root
getStateField(tmpState[], slot) == bs1_3.slot
# Move back slots, but not blocks
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.updateState(
tmpState[], dag.parent(bs1_3.bid).expect("block").atSlot(), false, cache)
tmpState[].latest_block_root == b1Add[].parent.root
getStateField(tmpState[], slot) == b1Add[].parent.slot
# Move to different block and slot
check:
dag.updateState(tmpState[], bs2_3, false, cache)
tmpState[].latest_block_root == b2Add[].root
getStateField(tmpState[], slot) == bs2_3.slot
# Move back slot and block
check:
dag.updateState(tmpState[], bs1, false, cache)
tmpState[].latest_block_root == b1Add[].root
getStateField(tmpState[], slot) == bs1.slot
# Move back to genesis
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.updateState(
tmpState[], dag.parent(bs1.bid).expect("block").atSlot(), false, cache)
tmpState[].latest_block_root == b1Add[].parent.root
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
getStateField(tmpState[], slot) == b1Add[].parent.slot
when declared(GC_fullCollect): # i386 test machines seem to run low..
GC_fullCollect()
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
suite "Block pool altair processing" & preset():
setup:
var
cfg = defaultRuntimeConfig
cfg.ALTAIR_FORK_EPOCH = Epoch(1)
var
db = makeTestDB(SLOTS_PER_EPOCH)
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, cfg, db, validatorMonitor, {})
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = Quarantine.init()
state = newClone(dag.headState)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
cache = StateCache()
info = ForkedEpochInfo()
# Advance to altair
check:
process_slots(
cfg, state[], cfg.ALTAIR_FORK_EPOCH.start_slot(), cache,
info, {}).isOk()
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
state[].kind == BeaconStateFork.Altair
var
b1 = addTestBlock(state[], cache).altairData
att1 = makeFullAttestations(state[], b1.root, b1.message.slot, cache)
b2 = addTestBlock(state[], cache, attestations = att1).altairData
test "Invalid signatures" & preset():
let badSignature = get_slot_signature(
Fork(), ZERO_HASH, 42.Slot,
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
MockPrivKeys[ValidatorIndex(0)]).toValidatorSig()
check:
dag.addHeadBlock(verifier, b1, nilAltairCallback).isOk()
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
block: # Main signature
var b = b2
b.signature = badSignature
let
bAdd = dag.addHeadBlock(verifier, b, nilAltairCallback)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
check:
bAdd.error() == BlockError.Invalid
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
block: # Randao reveal
var b = b2
b.message.body.randao_reveal = badSignature
let
bAdd = dag.addHeadBlock(verifier, b, nilAltairCallback)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
check:
bAdd.error() == BlockError.Invalid
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
block: # Attestations
var b = b2
b.message.body.attestations[0].signature = badSignature
let
bAdd = dag.addHeadBlock(verifier, b, nilAltairCallback)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
check:
bAdd.error() == BlockError.Invalid
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
block: # SyncAggregate empty
var b = b2
b.message.body.sync_aggregate.sync_committee_signature = badSignature
let
bAdd = dag.addHeadBlock(verifier, b, nilAltairCallback)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
check:
bAdd.error() == BlockError.Invalid
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
block: # SyncAggregate junk
var b = b2
b.message.body.sync_aggregate.sync_committee_signature = badSignature
b.message.body.sync_aggregate.sync_committee_bits[0] = true
let
bAdd = dag.addHeadBlock(verifier, b, nilAltairCallback)
Speed up altair block processing 2x (#3115) * Speed up altair block processing >2x Like #3089, this PR drastially speeds up historical REST queries and other long state replays. * cache sync committee validator indices * use ~80mb less memory for validator pubkey mappings * batch-verify sync aggregate signature (fixes #2985) * document sync committee hack with head block vs sync message block * add batch signature verification failure tests Before: ``` ../env.sh nim c -d:release -r ncli_db --db:mainnet_0/db bench --start-slot:-1000 All time are ms Average, StdDev, Min, Max, Samples, Test Validation is turned off meaning that no BLS operations are performed 5830.675, 0.000, 5830.675, 5830.675, 1, Initialize DB 0.481, 1.878, 0.215, 59.167, 981, Load block from database 8422.566, 0.000, 8422.566, 8422.566, 1, Load state from database 6.996, 1.678, 0.042, 14.385, 969, Advance slot, non-epoch 93.217, 8.318, 84.192, 122.209, 32, Advance slot, epoch 20.513, 23.665, 11.510, 201.561, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` After: ``` 7081.422, 0.000, 7081.422, 7081.422, 1, Initialize DB 0.553, 2.122, 0.175, 66.692, 981, Load block from database 5439.446, 0.000, 5439.446, 5439.446, 1, Load state from database 6.829, 1.575, 0.043, 12.156, 969, Advance slot, non-epoch 94.716, 2.749, 88.395, 100.026, 32, Advance slot, epoch 11.636, 23.766, 4.889, 205.250, 981, Apply block, no slot processing 0.000, 0.000, 0.000, 0.000, 0, Database load 0.000, 0.000, 0.000, 0.000, 0, Database store ``` * add comment
2021-11-24 12:43:50 +00:00
check:
bAdd.error() == BlockError.Invalid
suite "chain DAG finalization tests" & preset():
setup:
var
db = makeTestDB(SLOTS_PER_EPOCH)
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor, {})
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = Quarantine.init()
cache = StateCache()
info = ForkedEpochInfo()
test "prune heads on finalization" & preset():
# Create a fork that will not be taken
var
blck = makeTestBlock(dag.headState, cache).phase0Data
tmpState = assignClone(dag.headState)
check:
process_slots(
Implement split preset/config support (#2710) * Implement split preset/config support This is the initial bulk refactor to introduce runtime config values in a number of places, somewhat replacing the existing mechanism of loading network metadata. It still needs more work, this is the initial refactor that introduces runtime configuration in some of the places that need it. The PR changes the way presets and constants work, to match the spec. In particular, a "preset" now refers to the compile-time configuration while a "cfg" or "RuntimeConfig" is the dynamic part. A single binary can support either mainnet or minimal, but not both. Support for other presets has been removed completely (can be readded, in case there's need). There's a number of outstanding tasks: * `SECONDS_PER_SLOT` still needs fixing * loading custom runtime configs needs redoing * checking constants against YAML file * yeerongpilly support `build/nimbus_beacon_node --network=yeerongpilly --discv5:no --log-level=DEBUG` * load fork epoch from config * fix fork digest sent in status * nicer error string for request failures * fix tools * one more * fixup * fixup * fixup * use "standard" network definition folder in local testnet Files are loaded from their standard locations, including genesis etc, to conform to the format used in the `eth2-networks` repo. * fix launch scripts, allow unknown config values * fix base config of rest test * cleanups * bundle mainnet config using common loader * fix spec links and names * only include supported preset in binary * drop yeerongpilly, add altair-devnet-0, support boot_enr.yaml
2021-07-12 13:01:38 +00:00
defaultRuntimeConfig, tmpState[],
getStateField(tmpState[], slot) + (5 * SLOTS_PER_EPOCH).uint64,
cache, info, {}).isOk()
let lateBlock = addTestBlock(tmpState[], cache).phase0Data
block:
let status = dag.addHeadBlock(verifier, blck, nilPhase0Callback)
check: status.isOk()
assign(tmpState[], dag.headState)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
# skip slots so we can test gappy getBlockIdAtSlot
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
check process_slots(
defaultRuntimeConfig, tmpState[],
getStateField(tmpState[], slot) + 2.uint64,
cache, info, {}).isOk()
for i in 0 ..< (SLOTS_PER_EPOCH * 6):
if i == 1:
# There are 2 heads now because of the fork at slot 1
check:
dag.heads.len == 2
blck = addTestBlock(
tmpState[], cache,
attestations = makeFullAttestations(
tmpState[], dag.head.root, getStateField(tmpState[], slot), cache, {})).phase0Data
let added = dag.addHeadBlock(verifier, blck, nilPhase0Callback)
check: added.isOk()
dag.updateHead(added[], quarantine)
dag.pruneAtFinalization()
check:
dag.heads.len() == 1
dag.getBlockIdAtSlot(0.Slot).get().bid.slot == 0.Slot
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockIdAtSlot(2.Slot).get() ==
BlockSlotId.init(dag.getBlockIdAtSlot(1.Slot).get().bid, 2.Slot)
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockIdAtSlot(dag.head.slot).get() == BlockSlotId.init(
dag.head.bid, dag.head.slot)
dag.getBlockIdAtSlot(dag.head.slot + 1).get() == BlockSlotId.init(
dag.head.bid, dag.head.slot + 1)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
not dag.containsForkBlock(dag.getBlockIdAtSlot(5.Slot).get().bid.root)
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.containsForkBlock(dag.finalizedHead.blck.root)
dag.getBlockRef(dag.getBlockIdAtSlot(0.Slot).get().bid.root).isNone() # Finalized - no BlockRef
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockRef(dag.finalizedHead.blck.root).isSome()
isNil dag.finalizedHead.blck.parent
check:
dag.db.immutableValidators.len() == getStateField(dag.headState, validators).len()
block:
var cur = dag.head.bid
while true:
let parent = dag.parent(cur)
if cur.slot > 0:
check:
parent.isSome and parent.get().slot < cur.slot
cur = parent.get()
else:
check:
parent.isErr()
break
check: cur.slot == 0
block:
var cur = dag.head.bid.atSlot()
while true:
let parent = dag.parentOrSlot(cur)
if cur.slot > 0:
check:
parent.isSome and (parent.get().slot < cur.slot or parent.get().bid != cur.bid)
cur = parent.get()
else:
check:
parent.isErr()
break
check: cur.slot == 0
let
finalER = dag.getEpochRef(
dag.finalizedHead.blck, dag.finalizedHead.slot.epoch, false)
# The EpochRef for the finalized block is needed for eth1 voting, so we
# should never drop it!
check:
not finalER.isErr()
block:
for er in dag.epochRefs.entries:
check: er.value == nil or er.value.epoch >= dag.finalizedHead.slot.epoch
block:
let tmpStateData = assignClone(dag.headState)
# Check that cached data is available after updateState - since we
# just processed the head the relevant epochrefs should not have been
# evicted yet
cache = StateCache()
check: updateState(
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag, tmpStateData[],
dag.head.atSlot(dag.head.slot).toBlockSlotId().expect("not nil"),
false, cache)
check:
dag.head.slot.epoch in cache.shuffled_active_validator_indices
(dag.head.slot.epoch - 1) in cache.shuffled_active_validator_indices
dag.head.slot in cache.beacon_proposer_indices
block:
# The late block is a block whose parent was finalized long ago and thus
# is no longer a viable head candidate
let status = dag.addHeadBlock(verifier, lateBlock, nilPhase0Callback)
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
# This _should_ be Unviable, but we can't tell, from the data that we have
# so MissingParent is the least wrong thing to reply
check: status.error == BlockError.UnviableFork
block:
let
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
finalizedCheckpoint = dag.stateCheckpoint(dag.finalizedHead.toBlockSlotId().get())
headCheckpoint = dag.stateCheckpoint(dag.head.bid.atSlot())
prunedCheckpoint = dag.stateCheckpoint(dag.parent(dag.finalizedHead.blck.bid).get().atSlot())
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
db.getStateRoot(headCheckpoint.bid.root, headCheckpoint.slot).isSome
db.getStateRoot(finalizedCheckpoint.bid.root, finalizedCheckpoint.slot).isSome
db.getStateRoot(prunedCheckpoint.bid.root, prunedCheckpoint.slot).isNone
# Roll back head block (e.g., because it was declared INVALID)
let parentRoot = dag.head.parent.root
dag.updateHead(dag.head.parent, quarantine)
check: dag.head.root == parentRoot
let
validatorMonitor2 = newClone(ValidatorMonitor.init())
dag2 = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor2, {})
# check that the state reloaded from database resembles what we had before
check:
dag2.tail.root == dag.tail.root
dag2.head.root == dag.head.root
dag2.head.root == parentRoot
dag2.finalizedHead.blck.root == dag.finalizedHead.blck.root
dag2.finalizedHead.slot == dag.finalizedHead.slot
getStateRoot(dag2.headState) == getStateRoot(dag.headState)
# No canonical block data should be pruned by the removal of the fork
for i in Slot(0)..dag2.head.slot:
let bids = dag.getBlockIdAtSlot(i).expect("found it")
if bids.isProposed:
check: dag2.getForkedBlock(bids.bid).isSome
# The unviable block should have been pruned however
check: dag2.getForkedBlock(lateBlock.root).isNone
test "orphaned epoch block" & preset():
let prestate = (ref ForkedHashedBeaconState)(kind: BeaconStateFork.Phase0)
for i in 0 ..< SLOTS_PER_EPOCH:
if i == SLOTS_PER_EPOCH - 1:
assign(prestate[], dag.headState)
let blck = makeTestBlock(dag.headState, cache).phase0Data
let added = dag.addHeadBlock(verifier, blck, nilPhase0Callback)
check: added.isOk()
dag.updateHead(added[], quarantine)
dag.pruneAtFinalization()
check:
dag.heads.len() == 1
# The loop creates multiple branches, which StateCache isn't suitable for
cache = StateCache()
doAssert process_slots(
Implement split preset/config support (#2710) * Implement split preset/config support This is the initial bulk refactor to introduce runtime config values in a number of places, somewhat replacing the existing mechanism of loading network metadata. It still needs more work, this is the initial refactor that introduces runtime configuration in some of the places that need it. The PR changes the way presets and constants work, to match the spec. In particular, a "preset" now refers to the compile-time configuration while a "cfg" or "RuntimeConfig" is the dynamic part. A single binary can support either mainnet or minimal, but not both. Support for other presets has been removed completely (can be readded, in case there's need). There's a number of outstanding tasks: * `SECONDS_PER_SLOT` still needs fixing * loading custom runtime configs needs redoing * checking constants against YAML file * yeerongpilly support `build/nimbus_beacon_node --network=yeerongpilly --discv5:no --log-level=DEBUG` * load fork epoch from config * fix fork digest sent in status * nicer error string for request failures * fix tools * one more * fixup * fixup * fixup * use "standard" network definition folder in local testnet Files are loaded from their standard locations, including genesis etc, to conform to the format used in the `eth2-networks` repo. * fix launch scripts, allow unknown config values * fix base config of rest test * cleanups * bundle mainnet config using common loader * fix spec links and names * only include supported preset in binary * drop yeerongpilly, add altair-devnet-0, support boot_enr.yaml
2021-07-12 13:01:38 +00:00
defaultRuntimeConfig, prestate[], getStateField(prestate[], slot) + 1,
cache, info, {}).isOk()
# create another block, orphaning the head
let blck = makeTestBlock(prestate[], cache).phase0Data
# Add block, but don't update head
let added = dag.addHeadBlock(verifier, blck, nilPhase0Callback)
check: added.isOk()
var
validatorMonitor2 = newClone(ValidatorMonitor.init())
dag2 = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor2, {})
# check that we can apply the block after the orphaning
let added2 = dag2.addHeadBlock(verifier, blck, nilPhase0Callback)
check: added2.isOk()
test "init with gaps" & preset():
for blck in makeTestBlocks(
dag.headState, cache, int(SLOTS_PER_EPOCH * 6 - 2),
true):
let added = dag.addHeadBlock(verifier, blck.phase0Data, nilPhase0Callback)
check: added.isOk()
dag.updateHead(added[], quarantine)
dag.pruneAtFinalization()
# Advance past epoch so that the epoch transition is gapped
check:
process_slots(
defaultRuntimeConfig, dag.headState, Slot(SLOTS_PER_EPOCH * 6 + 2),
cache, info, {}).isOk()
let blck = makeTestBlock(
dag.headState, cache,
attestations = makeFullAttestations(
dag.headState, dag.head.root, getStateField(dag.headState, slot),
cache, {})).phase0Data
let added = dag.addHeadBlock(verifier, blck, nilPhase0Callback)
check: added.isOk()
dag.updateHead(added[], quarantine)
dag.pruneAtFinalization()
block:
# Check that we can rewind to every block from head to finalized
var
cur = dag.head
tmpStateData = assignClone(dag.headState)
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
while cur != nil: # Go all the way to dag.finalizedHead
assign(tmpStateData[], dag.headState)
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.updateState(tmpStateData[], cur.bid.atSlot(), false, cache)
dag.getForkedBlock(cur.bid).get().phase0Data.message.state_root ==
getStateRoot(tmpStateData[])
getStateRoot(tmpStateData[]) == hash_tree_root(
tmpStateData[].phase0Data.data)
cur = cur.parent
let
validatorMonitor2 = newClone(ValidatorMonitor.init())
dag2 = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor2, {})
# check that the state reloaded from database resembles what we had before
check:
dag2.tail.root == dag.tail.root
dag2.head.root == dag.head.root
dag2.finalizedHead.blck.root == dag.finalizedHead.blck.root
dag2.finalizedHead.slot == dag.finalizedHead.slot
getStateRoot(dag2.headState) == getStateRoot(dag.headState)
suite "Old database versions" & preset():
setup:
let
genState = newClone(initialize_hashed_beacon_state_from_eth1(
defaultRuntimeConfig, ZERO_HASH, 0,
makeInitialDeposits(SLOTS_PER_EPOCH.uint64, flags = {skipBlsValidation}),
{skipBlsValidation}))
genBlock = get_initial_beacon_block(genState[])
var
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = Quarantine.init()
test "pre-1.1.0":
# only kvstore, no immutable validator keys
let db = BeaconChainDB.new("", inMemory = true)
# preInit a database to a v1.0.12 state
db.putStateRoot(
genState[].latest_block_root, genState[].data.slot, genState[].root)
db.putStateV0(genState[].root, genState[].data)
db.putBlockV0(genBlock)
db.putTailBlock(genBlock.root)
db.putHeadBlock(genBlock.root)
db.putGenesisBlock(genBlock.root)
var
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db,validatorMonitor, {})
state = newClone(dag.headState)
cache = StateCache()
att0 = makeFullAttestations(state[], dag.tail.root, 0.Slot, cache)
b1 = addTestBlock(state[], cache, attestations = att0).phase0Data
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
check:
b1Add.isOk()
suite "Diverging hardforks":
setup:
var
phase0RuntimeConfig = defaultRuntimeConfig
altairRuntimeConfig = defaultRuntimeConfig
phase0RuntimeConfig.ALTAIR_FORK_EPOCH = FAR_FUTURE_EPOCH
altairRuntimeConfig.ALTAIR_FORK_EPOCH = 2.Epoch
var
db = makeTestDB(SLOTS_PER_EPOCH)
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, phase0RuntimeConfig, db, validatorMonitor, {})
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = newClone(Quarantine.init())
cache = StateCache()
info = ForkedEpochInfo()
tmpState = assignClone(dag.headState)
test "Tail block only in common":
check:
process_slots(
phase0RuntimeConfig, tmpState[],
getStateField(tmpState[], slot) + (3 * SLOTS_PER_EPOCH).uint64,
cache, info, {}).isOk()
# Because the first block is after the Altair transition, the only block in
# common is the tail block
var
b1 = addTestBlock(tmpState[], cache).phase0Data
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
check b1Add.isOk()
dag.updateHead(b1Add[], quarantine[])
let validatorMonitorAltair = newClone(ValidatorMonitor.init())
let dagAltair = init(
ChainDAGRef, altairRuntimeConfig, db, validatorMonitorAltair, {})
discard AttestationPool.init(dagAltair, quarantine)
test "Non-tail block in common":
check:
process_slots(
phase0RuntimeConfig, tmpState[],
getStateField(tmpState[], slot) + SLOTS_PER_EPOCH.uint64,
cache, info, {}).isOk()
# There's a block in the shared-correct phase0 hardfork, before epoch 2
var
b1 = addTestBlock(tmpState[], cache).phase0Data
b1Add = dag.addHeadBlock(verifier, b1, nilPhase0Callback)
check:
b1Add.isOk()
process_slots(
phase0RuntimeConfig, tmpState[],
getStateField(tmpState[], slot) + (3 * SLOTS_PER_EPOCH).uint64,
cache, info, {}).isOk()
var
b2 = addTestBlock(tmpState[], cache).phase0Data
b2Add = dag.addHeadBlock(verifier, b2, nilPhase0Callback)
check b2Add.isOk()
dag.updateHead(b2Add[], quarantine[])
let validatorMonitor = newClone(ValidatorMonitor.init())
let dagAltair = init(
ChainDAGRef, altairRuntimeConfig, db, validatorMonitor, {})
discard AttestationPool.init(dagAltair, quarantine)
suite "Backfill":
setup:
let
genState = (ref ForkedHashedBeaconState)(
kind: BeaconStateFork.Phase0,
phase0Data: initialize_hashed_beacon_state_from_eth1(
defaultRuntimeConfig, ZERO_HASH, 0,
makeInitialDeposits(SLOTS_PER_EPOCH.uint64, flags = {skipBlsValidation}),
{skipBlsValidation}))
tailState = assignClone(genState[])
blocks = block:
var blocks: seq[ForkedSignedBeaconBlock]
var cache: StateCache
for i in 0..<SLOTS_PER_EPOCH * 2:
blocks.add addTestBlock(tailState[], cache)
blocks
let
db = BeaconChainDB.new("", inMemory = true)
test "backfill to genesis":
let
tailBlock = blocks[^1]
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
genBlock = get_initial_beacon_block(genState[])
ChainDAGRef.preInit(
db, genState[], tailState[], tailBlock.asTrusted())
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor, {})
var cache = StateCache()
check:
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockRef(tailBlock.root).get().bid == dag.tail
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.getBlockRef(blocks[^2].root).isNone()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockId(tailBlock.root).get() == dag.tail
dag.getBlockId(blocks[^2].root).isNone()
dag.getBlockIdAtSlot(dag.tail.slot).get().bid == dag.tail
dag.getBlockIdAtSlot(dag.tail.slot - 1).isNone()
dag.getBlockIdAtSlot(Slot(0)).isSome() # genesis stored in db
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockIdAtSlot(Slot(1)).isNone()
# No EpochRef for pre-tail epochs
dag.getEpochRef(dag.tail, dag.tail.slot.epoch - 1, true).isErr()
# Should get EpochRef for the tail however
dag.getEpochRef(dag.tail, dag.tail.slot.epoch, true).isOk()
dag.getEpochRef(dag.tail, dag.tail.slot.epoch + 1, true).isOk()
# Should not get EpochRef for random block
dag.getEpochRef(
BlockId(root: blocks[^2].root, slot: dag.tail.slot), # root/slot mismatch
dag.tail.slot.epoch, true).isErr()
dag.getEpochRef(dag.tail, dag.tail.slot.epoch + 1, true).isOk()
dag.getFinalizedEpochRef() != nil
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
dag.backfill == tailBlock.phase0Data.message.toBeaconBlockSummary()
# Check that we can propose right from the checkpoint state
dag.getProposalState(dag.head, dag.head.slot + 1, cache).isOk()
var
badBlock = blocks[^2].phase0Data
badBlock.signature = blocks[^3].phase0Data.signature
check:
dag.addBackfillBlock(badBlock).error == BlockError.Invalid
check:
dag.addBackfillBlock(blocks[^3].phase0Data).error == BlockError.MissingParent
dag.addBackfillBlock(tailBlock.phase0Data).error == BlockError.Duplicate
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.addBackfillBlock(genBlock.phase0Data.asSigned()).error == BlockError.MissingParent
check:
dag.addBackfillBlock(blocks[^2].phase0Data).isOk()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockRef(tailBlock.root).get().bid == dag.tail
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag.getBlockRef(blocks[^2].root).isNone()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockId(tailBlock.root).get() == dag.tail
dag.getBlockId(blocks[^2].root).get().root == blocks[^2].root
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag.getBlockIdAtSlot(dag.tail.slot).get().bid == dag.tail
dag.getBlockIdAtSlot(dag.tail.slot - 1).get() ==
blocks[^2].toBlockId().atSlot()
dag.getBlockIdAtSlot(dag.tail.slot - 2).isNone
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
dag.backfill == blocks[^2].phase0Data.message.toBeaconBlockSummary()
check:
dag.addBackfillBlock(blocks[^3].phase0Data).isOk()
dag.getBlockIdAtSlot(dag.tail.slot - 2).get() ==
blocks[^3].toBlockId().atSlot()
dag.getBlockIdAtSlot(dag.tail.slot - 3).isNone
for i in 3..<blocks.len:
check: dag.addBackfillBlock(blocks[blocks.len - i - 1].phase0Data).isOk()
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
check:
dag.addBackfillBlock(genBlock.phase0Data.asSigned).error == BlockError.Duplicate
dag.backfill.slot == GENESIS_SLOT
dag.rebuildIndex()
check:
dag.getFinalizedEpochRef() != nil
test "reload backfill position":
let
tailBlock = blocks[^1]
ChainDAGRef.preInit(
db, genState[], tailState[], tailBlock.asTrusted())
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor, {})
check:
dag.addBackfillBlock(blocks[^2].phase0Data).isOk()
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
dag.backfill == blocks[^2].phase0Data.message.toBeaconBlockSummary()
let
validatorMonitor2 = newClone(ValidatorMonitor.init())
dag2 = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor2, {})
check:
dag2.getFinalizedEpochRef() != nil
limit by-root requests to non-finalized blocks (#3293) * limit by-root requests to non-finalized blocks Presently, we keep a mapping from block root to `BlockRef` in memory - this has simplified reasoning about the dag, but is not sustainable with the chain growing. We can distinguish between two cases where by-root access is useful: * unfinalized blocks - this is where the beacon chain is operating generally, by validating incoming data as interesting for future fork choice decisions - bounded by the length of the unfinalized period * finalized blocks - historical access in the REST API etc - no bounds, really In this PR, we limit the by-root block index to the first use case: finalized chain data can more efficiently be addressed by slot number. Future work includes: * limiting the `BlockRef` horizon in general - each instance is 40 bytes+overhead which adds up - this needs further refactoring to deal with the tail vs state problem * persisting the finalized slot-to-hash index - this one also keeps growing unbounded (albeit slowly) Anyway, this PR easily shaves ~128mb of memory usage at the time of writing. * No longer honor `BeaconBlocksByRoot` requests outside of the non-finalized period - previously, Nimbus would generously return any block through this libp2p request - per the spec, finalized blocks should be fetched via `BeaconBlocksByRange` instead. * return `Opt[BlockRef]` instead of `nil` when blocks can't be found - this becomes a lot more common now and thus deserves more attention * `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized blocks from now - `finalizedBlocks` covers the other `BlockRef` instances * in backfill, verify that the last backfilled block leads back to genesis, or panic * add backfill timings to log * fix missing check that `BlockRef` block can be fetched with `getForkedBlock` reliably * shortcut doppelganger check when feature is not enabled * in REST/JSON-RPC, fetch blocks without involving `BlockRef` * fix dag.blocks ref
2022-01-21 11:33:16 +00:00
dag2.getBlockRef(tailBlock.root).get().root == dag.tail.root
dag2.getBlockRef(blocks[^2].root).isNone()
Prune `BlockRef` on finalization (#3513) Up til now, the block dag has been using `BlockRef`, a structure adapted for a full DAG, to represent all of chain history. This is a correct and simple design, but does not exploit the linearity of the chain once parts of it finalize. By pruning the in-memory `BlockRef` structure at finalization, we save, at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory landing us at a steady state of ~750mb normal memory usage for a validating node. Above all though, we prevent memory usage from growing proportionally with the length of the chain, something that would not be sustainable over time - instead, the steady state memory usage is roughly determined by the validator set size which grows much more slowly. With these changes, the core should remain sustainable memory-wise post-merge all the way to withdrawals (when the validator set is expected to grow). In-memory indices are still used for the "hot" unfinalized portion of the chain - this ensure that consensus performance remains unchanged. What changes is that for historical access, we use a db-based linear slot index which is cache-and-disk-friendly, keeping the cost for accessing historical data at a similar level as before, achieving the savings at no percievable cost to functionality or performance. A nice collateral benefit is the almost-instant startup since we no longer load any large indicies at dag init. The cost of this functionality instead can be found in the complexity of having to deal with two ways of traversing the chain - by `BlockRef` and by slot. * use `BlockId` instead of `BlockRef` where finalized / historical data may be required * simplify clearance pre-advancement * remove dag.finalizedBlocks (~50:ish mb) * remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead * `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef` instance, unlike `BlockRef` traversal * prune `BlockRef` parents on finality (~200:ish mb) * speed up ChainDAG init by not loading finalized history index * mess up light client server error handling - this need revisiting :)
2022-03-17 17:42:56 +00:00
dag2.getBlockIdAtSlot(dag.tail.slot).get().bid.root == dag.tail.root
dag2.getBlockIdAtSlot(dag.tail.slot - 1).get() ==
blocks[^2].toBlockId().atSlot()
dag2.getBlockIdAtSlot(dag.tail.slot - 2).isNone
chaindag: don't keep backfill block table in memory (#3429) This PR names and documents the concept of the archive: a range of slots for which we have degraded functionality in terms of historical access - in particular: * we don't support rewinding to states in this range * we don't keep an in-memory representation of the block dag The archive de-facto exists in a trusted-node-synced node, but this PR gives it a name and drops the in-memory digest index. In order to satisfy `GetBlocksByRange` requests, we ensure that we have blocks for the entire archive period via backfill. Future versions may relax this further, adding a "pre-archive" period that is fully pruned. During by-slot searches in the archive (both for libp2p and rest requests), an extra database lookup is used to covert the given `slot` to a `root` - future versions will avoid this using era files which natively are indexed by `slot`. That said, the lookup is quite fast compared to the actual block loading given how trivial the table is - it's hard to measure, even. A collateral benefit of this PR is that checkpoint-synced nodes will see 100-200MB memory usage savings, thanks to the dropped in-memory cache - future pruning work will bring this benefit to full nodes as well. * document chaindag storage architecture and assumptions * look up parent using block id instead of full block in clearance (future-proofing the code against a future in which blocks come from era files) * simplify finalized block init, always writing the backfill portion to db at startup (to ensure lookups work as expected) * preallocate some extra memory for finalized blocks, to avoid immediate realloc
2022-02-26 18:16:19 +00:00
dag2.backfill == blocks[^2].phase0Data.message.toBeaconBlockSummary()
test "Init without genesis / block":
let
tailBlock = blocks[^1]
genBlock = get_initial_beacon_block(genState[])
ChainDAGRef.preInit(db, tailState[])
let
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor, {})
check:
dag.getFinalizedEpochRef() != nil
for i in 0..<blocks.len:
check: dag.addBackfillBlock(
blocks[blocks.len - i - 1].phase0Data).isOk()
check:
dag.addBackfillBlock(genBlock.phase0Data.asSigned).isOk()
dag.addBackfillBlock(
genBlock.phase0Data.asSigned).error == BlockError.Duplicate
var
cache: StateCache
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = newClone(Quarantine.init())
let
next = addTestBlock(tailState[], cache).phase0Data
nextAdd = dag.addHeadBlock(verifier, next, nilPhase0Callback).get()
dag.updateHead(nextAdd, quarantine[])
let
validatorMonitor2 = newClone(ValidatorMonitor.init())
dag2 = init(ChainDAGRef, defaultRuntimeConfig, db, validatorMonitor2, {})
check:
dag2.head.root == next.root
suite "Latest valid hash" & preset():
setup:
var runtimeConfig = defaultRuntimeConfig
runtimeConfig.ALTAIR_FORK_EPOCH = 1.Epoch
runtimeConfig.BELLATRIX_FORK_EPOCH = 2.Epoch
var
db = makeTestDB(SLOTS_PER_EPOCH)
validatorMonitor = newClone(ValidatorMonitor.init())
dag = init(ChainDAGRef, runtimeConfig, db, validatorMonitor, {})
verifier = BatchVerifier(rng: keys.newRng(), taskpool: Taskpool.new())
quarantine = newClone(Quarantine.init())
cache = StateCache()
info = ForkedEpochInfo()
state = newClone(dag.headState)
test "LVH searching":
# Reach Bellatrix, where execution payloads exist
check process_slots(
runtimeConfig, state[],
getStateField(state[], slot) + (3 * SLOTS_PER_EPOCH).uint64,
cache, info, {}).isOk()
var
b1 = addTestBlock(state[], cache, cfg = runtimeConfig).bellatrixData
b1Add = dag.addHeadBlock(verifier, b1, nilBellatrixCallback)
b2 = addTestBlock(state[], cache, cfg = runtimeConfig).bellatrixData
b2Add = dag.addHeadBlock(verifier, b2, nilBellatrixCallback)
b3 = addTestBlock(state[], cache, cfg = runtimeConfig).bellatrixData
b3Add = dag.addHeadBlock(verifier, b3, nilBellatrixCallback)
dag.updateHead(b3Add[], quarantine[])
check: dag.head.root == b3.root
# Ensure that head can go backwards in case of head being marked invalid
dag.updateHead(b2Add[], quarantine[])
check: dag.head.root == b2.root
dag.updateHead(b1Add[], quarantine[])
check: dag.head.root == b1.root
const fallbackEarliestInvalid =
Eth2Digest.fromHex("0x0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef")
check:
# Represents where LVH is two behind the invalid-marked block (because
# first param is parent). It searches using LVH (i.e. execution hash),
# but returns CL block hash, because that's what fork choice and other
# Nimbus components mostly use as a coordinate system. Since b1 is set
# to be valid here by being the LVH, it means that b2 must be invalid.
dag.getEarliestInvalidBlockRoot(
b2Add[].root, b1.message.body.execution_payload.block_hash,
fallbackEarliestInvalid) == b2Add[].root
# This simulates calling it based on b3 (child of b2), where there's no
# gap in detecting the invalid blocks. Because the API, due to testcase
# design, does not assume the block being tested is in the DAG, there's
# a manually specified fallback (CL) block root to use, because it does
# not have access to this information otherwise, because the very first
# newest block in the chain it's examining is already valid.
dag.getEarliestInvalidBlockRoot(
b2Add[].root, b2.message.body.execution_payload.block_hash,
fallbackEarliestInvalid) == fallbackEarliestInvalid