When EL `newPayload` is slow (e.g., Raspberry Pi with Besu), the epoch
and shuffling caches tend to fill up with multiple copies per epoch when
processing gossip and performing validator duties close to wall slot.
The old strategy of evicting oldest epoch led to the same item being
evicted over and over, leading to blocking of over 5 minutes in extreme
cases where alternate epochs/shuffling got loaded repeatedly.
Changing the cache eviction strategy to least-recently-used seems to
improve the situation drastically. A simple implementation was selected
based on single linked-list without a hashtable.
When backfilling LC updates (`--light-client-data-import-mode=full`),
the highest participation update is computed without ensuring that the
finalized header is in the same period. Updates sharing same period for
both finalized and attested headers should be preferred.
Fixes a bug leading to suboptimal update selection.
* avoid database race-condition inconsistency after fcU `INVALID` then crash
* ensure head doesn't fall behind finalized; add more tests for head movement/reloading DAG
In `eth2_rest_serialization` there was a `dump` function for
`KeystoresAndSlashingProtection` that does not seem to be used.
Removes that unused function.
For pre-encoded JSON REST responses we have `jsonResponsePlain`.
Adds a `sszResponsePlain` function to serve similar purpose for SSZ.
This avoids caller having to explicitly specify Http200 and media type.
When the BN's head is reorged while shut down, reloading the BN will not
assign `BlockRef` to alternate branches. However, blocks from other
branches are still present in the database, leading to their descendants
incorrectly marked as `UnviableFork`. By restricting the check to blocks
that have been finalized, they should be reported as `MissingParent`
instead, eventually re-assigning a `BlockRef` to them.
The `eth1_monitor` check to require engine API from bellatrix onward
has issues in setups where the EL and CL are started simultaneously
because the EL may not be ready to answer requests by the time that the
check is performed. This can be observed, e.g., on Raspberry Pi 4 when
using Besu as the EL client. Now that the merge transition happened, the
check is also not that useful anymore, as users have other ways to know
that their setup is not working correctly (e.g., repeated exchange logs)
When running as Gnosis-chain binary the config was no longer adjustable.
Restores loading custom configs when running as Gnosis-chain binary,
as long as the following keys remain same:
- SLOTS_PER_EPOCH=16
- SECONDS_PER_SLOT=5
- BASE_REWARD_FACTOR=25
- EPOCHS_PER_SYNC_COMMITTEE_PERIOD=512
This allows running the Gnosis-chain binary on custom test networks.
* detect mismatch of config and binary
When loading configuration that sets keys that Nimbus bakes into the
binary at compile-time, raise an error if the config is incompatible
instead of ignoring the conflicting value.
When TTD hits before Bellatrix, avoid waiting for new blocks and detect
the TTD block as the terminal block hash even before Bellatrix hits.
Also allow detecting EL genesis block as merge transition block.
This fixes the local testnet simulation with Geth to actually merge.
Since these files may have been created in a previous run or manually,
we want to keep loading them even on nodes that don't enable the
keystore API (for example static setups)
Other changes:
* log keystore loading progressively (#3699)
* print initial fee recipient when loading validators
* log dynamic fee recipient updates
The optimistic candidate block check that only imports a new block into
the EL client if its parent block also had execution enabled is not
needed anymore, as mainnet has merged and the attack period is over.
Port changes to `nextExchangeTransitionConfiguration` from BN to LC:
- 60 seconds delay before initial exchange
- 45 seconds interval between followup exchanges
- Only exchange post Bellatrix
When the sync queue processes results for a blocks by range request,
and the requested range contained some slots that are already finalized,
`BlockError.MissingParent` currently leads to `PeerScoreBadBlocks` even
when the error occurs on a non-finalized slot in the requested range.
This patch changes the scoring in that case to `PeerScoreMissingBlocks`
for consistency with range requests solely covering non-finalized slots,
and, likewise, rewinds the sync queue to the next `rewindSlot`.
Per spec, we should not be sending our detected terminal block to EL -
the EL configuration exchange should only look at values from
configuration and report mismatches.
`p.dataProvider` may become `nil` between individual attempts to
exchange transition configuration with the EL. Harden by capturing
the data provider on function start.
Note that other functions are already hardened, or are unaffected.
Only `close` transitions `p.dataProvider` to `nil`, and `close` is
only called by the main deposits import sequence. During the deposits
import, `close` is not called, so extra checks are not needed.
When there are a lot of deposits, we decompress the public key into a
crypto cache. To avoid having those caches grow unreasonably big,
make sure to operate on the decompressed pubkey instead.
* more efficient forkchoiceUpdated usage
* await rather than asyncSpawn; ensure head update before dag.updateHead
* use action tracker rather than attached validators to check for next slot proposal; use wall slot + 1 rather than state slot + 1 to correctly check when missing blocks
* re-add two-fcU case for when newPayload not VALID
* check dynamicFeeRecipientsStore for potential proposal
* remove duplicate checks for whether next proposer
`news` has a few open issues that are not present in `nim-websock`:
1. There is a 1 second delay between each MB of sent data.
2. Cancelling an ongoing `send` makes the entire WebSocket unusable.
3. Control packets do not have priority over ongoing message frames.
Using `news`, there are quite a few of these messages in Geth:
```
Previously seen beacon client is offline. Please ensure it is
operational to follow the chain!
```
It may take quite some time to reconnect when this happens.
Using `nim-websock`, this message still occurs because `eth1_monitor`
reconnects the EL connection when no new blocks occurred for 5 minutes,
but reconnecting is quick and the message is rarer.
The sync protocol does not distinguish between:
- All requested slots are empty
- Peer does not have data available about requested range
Therefore, we treat EOF for `beacon_blocks_by_range` and for
`beacon_blocks_by_range` as valid responses, as if the entire epoch
really contained no single block for any slot. Once a followup response
provides new blocks, we detect that some blocks were missing and rewind.
During backfill, we also request the known-to-exist `backfill.slot`,
so we can actually detect whether an epoch really does not have blocks
or whether a response is incomplete (`PeerScoreNoBlocks`).
When the BN-embedded LC makes sync progress, pass the corresponding
execution block hash to the EL via `engine_forkchoiceUpdatedV1`.
This allows the EL to sync to wall slot while the chain DAG is behind.
Renamed `--light-client` to `--sync-light-client` for clarity, and
`--light-client-trusted-block-root` to `--trusted-block-root` for
consistency with `nimbus_light_client`.
Note that this does not work well in practice at this time:
- Geth sticks to the optimistic sync:
"Ignoring payload while snap syncing" (when passing the LC head)
"Forkchoice requested unknown head" (when updating to LC head)
- Nethermind syncs to LC head but does not report ancestors as VALID,
so the main forward sync is still stuck in optimistic mode:
"Pre-pivot block, ignored and returned Syncing"
To aid EL client teams in fixing those issues, having this available
as a hidden option is still useful.
When issuing an engine API call while the EL is disconnected, a `nil`
pointer is dereferenced. Fixed by correctly initializing futures.
```
Traceback (most recent call last, using override)
vendor/nim-libp2p/libp2p/protocols/pubsub/pubsub.nim(890) main
beacon_chain/nimbus_beacon_node.nim(2139) main
beacon_chain/nimbus_beacon_node.nim(0) handleStartUpCmd
beacon_chain/nimbus_beacon_node.nim(0) doRunBeaconNode
beacon_chain/nimbus_beacon_node.nim(0) start
beacon_chain/nimbus_beacon_node.nim(1589) run
vendor/nimbus-build-system/vendor/Nim/lib/system/iterators_1.nim(107) poll
vendor/nim-chronos/chronos/asyncfutures2.nim(365) futureContinue
beacon_chain/consensus_object_pools/consensus_manager.nim(297) updateHeadWithExecution
vendor/nim-chronos/chronos/asyncmacro2.nim(213) runProposalForkchoiceUpdated
vendor/nim-chronos/chronos/asyncfutures2.nim(365) futureContinue
beacon_chain/consensus_object_pools/consensus_manager.nim(259) runProposalForkchoiceUpdated
beacon_chain/eth1/eth1_monitor.nim(0) forkchoiceUpdated
vendor/nim-chronos/chronos/asyncfutures2.nim(219) complete
vendor/nim-chronos/chronos/asyncfutures2.nim(149) cancelled
vendor/nimbus-build-system/vendor/Nim/lib/system/excpt.nim(610) signalHandler
SIGSEGV: Illegal storage access. (Attempt to read from nil?)
```
The optimistic sync spec was updated since the LC based optsync module
was introduced. It is no longer necessary to wait for the justified
checkpoint to have execution enabled; instead, any block is okay to be
optimistically imported to the EL client, as long as its parent block
has execution enabled. Complex syncing logic has been removed, and the
LC optsync module will now follow gossip directly, reducing the latency
when using this module. Note that because this is now based on gossip
instead of using sync manager / request manager, that individual blocks
may be missed. However, EL clients should recover from this by fetching
missing blocks themselves.
When the EL fails to respond to `newPayload`, e.g., because connection
to the EL got interrupted, or due to misconfiguration, optimistic blocks
cannot be imported according to spec. This condition is treated the same
as if the peer returned a block with missing parent which gets the block
out of our processing queue, but can have nasty side effects.
For example, if sync manager asks for validation of a block known to be
in the finalized range, if it receives a `MissingParent` verdict, the
peer is immediately removed from the peer pool.
```
DBG 2022-08-24 11:45:26.874+02:00 newPayload: inserting block into execution engine parentHash=e4ca7424 blockHash=36cdc198 stateRoot=cf3902c1 receiptsRoot=56e81f17 prevRandao=0b49a172 blockNumber=1518089 gasLimit=30000000 gasUsed=0 timestamp=1657980396 extraDataLen=0 baseFeePerGas=7 numTransactions=0
ERR 2022-08-24 11:45:26.875+02:00 newPayload failed msg="Transport is not initialised (missing a call to connect?)"
DBG 2022-08-24 11:45:26.875+02:00 Block pool rejected peer's response topics="syncman" request=187232:32@1475 peer=16U*MsCJdx direction=forward blocks_map=xxxxxxxxxxxxxxxxxxxxxxxxxxxx.xxx blocks_count=31 ok=false unviable=false missing_parent=true sync_ident=main
ERR 2022-08-24 11:45:26.875+02:00 Unexpected missing parent at finalized epoch slot topics="syncman" request=187232:32@1475 peer=16U*MsCJdx direction=forward rewind_to_slot=187232 blocks_count=31 blocks_map=xxxxxxxxxxxxxxxxxxxxxxxxxxxx.xxx sync_ident=main
DBG 2022-08-24 11:45:26.875+02:00 Peer was removed from PeerPool due to low score topics="beacnde" peer=16U*MsCJdx peer_score=-1000 score_low_limit=0 score_high_limit=1000
DBG 2022-08-24 11:45:26.875+02:00 Lost connection to peer topics="networking" peer=16U*MsCJdx connections=0
```
By delaying issuing a verdict until the EL connection is restored and
`newPayload` successfully ran, the problem should be fixed. This also
induces back pressure to the sync manager by stopping download of new
blocks (or re-downloading the same block over and over again).
* Harden block proposal against expired slashings/exits
When a message is signed in a phase0 domain, it can no longer be
validated under bellatrix due to the correct fork no longer being
available in the `BeaconState`.
To ensure that all slashing/exits are still valid, in this PR we re-run
the checks in the state that we're proposing for, thus hardening against
both signatures and other changes in the state that might have
invalidated the message.
* fix same message added multiple times
in case of attestation slashing of multiple validators in one go
* support connecting to peers without bellatrix
Make discovery fork ID aware of scheduled Bellatrix fork to enable
connections to peers that don't have Bellatrix scheduled yet.
Without this, has peering issues with peers on older SW version.
* expand tests with compatibility checks
* more exhaustive compatibility checks
Aligns the default retention policy for LC data with the one for blocks.
Minimum spec requirement for both blocks and LC data is ~5 months.
Additional use cases are better supported by retaining data for longer.
* Fixes a segfault during block production when the Keymanager API
is disabled. The Keymanager is now disabled on half of the local
testnet nodes to catch such problems in the future.
* Fixes multiple potential stalls from REST requests being done
without a timeout. From practice, we know that such requests
can hang forever if not cancelled with a timeout. At best,
this would be a resource leak, at worst, it may lead to a
full stall of the client and missed validator duties.
* Changes some Options usages to Opt (for easier use of valueOr)
When the client was started without any validators, the doppelganger
detection structures were never initialized properly. Later, when
validators were added through the Keymanager API, they interacted
with the uninitialized doppelganger detection structures and their
duties were inappropriately skipped.
* Keymanager API for the validator client
* Properly treat the 'description' field as optional when loading Keystores
* Spec-compliant serialization of the slashing data in Keymanager's DeleteKeys response ()
Fixes#3940Fixes#3964Closes#3884 by adding test
In order to avoid full replays when validating attestations hailing from
untaken forks, it's better to keep shufflings separate from `EpochRef`
and perform a lookahead on the shuffling when processing the block that
determines them.
This also helps performance in the case where REST clients are trying to
perform lookahead on attestation duties and decreases memory usage by
sharing shufflings between EpochRef instances of the same dependent
root.
* packaging updates
* one package per binary (nimbus_beacon_node, nimbus_validator_client)
* use `-` in package name (`_` is separating the version)
* don't include (un)installation scripts in package
* default metrics port 8108 for vc
* fix several upgrade/install errors in scripts
* add JWT option to service files
* don't attempt to remove user on purge
* import EL deposits even when EL is stuck
The `eth1_monitor` only starts importing deposits once the EL reports a
new head block. However, the EL may be stuck at a block, e.g., the TTD.
By polling the latest EL block once after subscribing to new EL block
events it is ensured that deposits are still imported in this situation.
* also poll once on re-connects
* update `eth1_latest_head` metric in poll mode
* add comment about similar polling vs events parts
* replace check with assert
* `isNewLastBlock` helper
When fetching eth1 data and deposits for a new block proposal, the list
of deposits from previous eth1 data to the next one is fully loaded into
a `seq`. This can potentially be a very long list in active periods.
Changing this to an `iterator` saves memory by ensuring that the entire
list is no longer materialized; only the `DepositData` roots are needed.
When the EL connection is interrupted, deposits are once more requested
in chunks of 5000 blocks. This is a problem when the response takes over
a minute to produce and consistently times out as followup requests with
lower chunk sizes may no longer work after a request was canceled, e.g.,
when using Geth with websockets. By keeping track of `blocksPerRequest`
across EL reconnections, it is possible to recover from this by avoiding
to continuously repeat the initial request with the full 5000 blocks.
Also cleans up one more "retry of retry" instance; `DataProviderTimeout`
is a `CatchableError` and already handled by the existing retry logic.
When connection to the EL is lost as part of EL deposits importing, the
targeted block range to sync would reset. This is changed to properly
remember import progress across reconnects.
https://github.com/status-im/nimbus-eth2/pull/3944
The use of nested `awaitWithRetries` calls would have
resulted in an unexpected number of retries (3x3).
We now use regular `await` in outer layer to avoid the problem.
https://github.com/status-im/nimbus-eth2/pull/3943
The new code has an invariant that the `headMerkleizer` field in
the `Eth1Chain` is always kept in sync with the blocks stored in
the chain.
This invariant is now enforced better by doing the necessary merkleizer updates
in the `Eth1Chain.addBlock` function, in the `Eth1Chain.init` function and in the
`Eth1Chain.reset` function.
When importing blocks with deposits from the EL, the timestamp is never
initialized for them. Therefore, only blocks without deposits (for which
the timestamp is obtained) are considered for `is_candidate_block`.
This is fixed by also importing timestamps for blocks with deposits.
* fix obtaining deposits after connection loss
When an error occurs during Eth1 deposits import, the already imported
blocks are kept while the connection to the EL is re-established.
However, the corresponding merkleizer is not persisted, leading to any
future deposits no longer being properly imported. This is quite common
when syncing a fresh Nimbus instance against an already-synced Geth EL.
Fixed by persisting the head merkleizer together with the blocks.
* MEV validator registration
* add nearby canary to detect new beacon chain forks
* remove special MEV graffiti
* web3signer support
* fix trace logging
* Nim 1.2 needs raises Defect
* use template rather than proc in REST JSON parsing
* use --payload-builder-enable and --payload-builder-url
* explicitly default MEV to disabled
* explicitly empty default value for payload builder URL
* revert attestation pool to unstable version
* Use final `v1` version for light client protocols
* Unhide LC data collection options
* Default enable LC data serving
* rm unneeded import
* Connect to EL on startup
* Add docs for LC based EL sync
LC cancels concurrent requests if one peer sent a correct response and
waiting for other peers is no longer useful. On the server side this
resulted in a descore (-500) and a likely disconnect. The behaviour is
changed to match `UnexpectedEOF`, `PotentiallyExpectedEOF` handling that
return an error response without disconnecting from the peer.
When there is heavy forking, proposals may get missed due to including
attestations from different forks that later fail verification.
Checking attestation signatures when building blocks should fix this.
The light client sync protocol employs heuristics to ensure it does not
become stuck during non-finality or low sync committee participation.
These can enable use cases that prefer availability of recent data
over security. For our syncing use case, though, security is preferred.
An option is added to light client processor to configure this tradeoff.
* Re-enabled requireAllFields after a fix in nim-json-serialization
The problem was that `Option[T]` fields were not treated as optional
when requireAllFields is set to true. This is now fixed in NJS.
* Add makefile targets for recreating the Jenkins simulation runs
* Fix a discrepancy with the REST spec
It's not quite clear why this condition was triggered in the local
simulation, but it seems a viable scenario after the Keymanager API
is integrated in the validator client.
The user can temporarily remove all validator keys from a running
client before adding another set of keys.
Adds the `--web3-url` launch argument to `nimbus_light_client` to enable
driving the EL with the optimistic head obtained from LC sync protocol.
This will keep issuing `newPayload` / `forkChoiceUpdated` requests for
new blocks, marking them as optimistic. `ZERO_HASH` is reported as the
finalized block for now.
#3864 introduced a regression by turning on `requireAllFields` globally
for JSON parsing. Certain endpoints such as `RestSyncInfo` have optional
fields that do not parse correctly without additional changes. This is
reverted for now to restore previous behaviour and unblock CI testing.
Other changes:
* The Keymanager error responses differ from the Beacon API responses.
'keymanagerApiError' replaces the former usages of 'jsonError'.
* Return status code 401 and 403 for authorization errors in accordance
to the spec.
* Eliminate inconsistencies in the REST JSON parsing. Some of the code
paths allowed missing fields.
* Added logging of serialization failure details at DEBUG level.
* track the SyncCommittee period in slot end logs
* Update beacon_chain/nimbus_beacon_node.nim
Co-authored-by: Etan Kissling <etan@status.im>
Co-authored-by: Etan Kissling <etan@status.im>
Removes a few extra-ambitious templates to make `self` updates explicit,
and moves the `FinalityCheckpoints` type from `base` to `helpers` as it
is an additional Nimbus specific type not defined by spec.
a notice in the log is enough - we don't want the REST API to return an
error in this case because that makes the validator client think
something is seriously wrong (like the BN or message being broken)
Whether new blocks/attestations/etc are produced internally or received
via REST, their journey through the node is the same - to ensure that
they get the same treatment (logging, metrics, processing), this PR
moves the routing to a dedicated module and fixes several small
differences that existed before.
* `xxxValidator` -> `processMessageName` - the processor also was adding
messages to pools, so we want the name to reflect that action
* add missing "sent" metrics for some messages
* document ignore policy better - already-seen messages are not actaully
rebroadcast by libp2p
* skip redundant signature checks for internal validators consistently
The justified and finalized `Checkpoint` are frequently passed around
together. This introduces a new `FinalityCheckpoint` data structure that
combines them into one.
Due to the large usage of this structure in fork choice, also took this
opportunity to update fork choice tests to the latest v1.2.0-rc.1 spec.
Many additional tests enabled, some need more work, e.g. EL mock blocks.
Also implemented `discard_equivocations` which was skipped in #3661,
and improved code reuse across fork choice logic while at it.
* merge LC db into main BN db
To treat derived LC data similar to derived state caches, merge it into
the main beacon node DB.
* shorten table names, group with lc prefix
* optimistic sync
* flag that initially loaded blocks from database might need execution block root filled in
* return optimistic status in REST calls
* refactor blockslot pruning
* ensure beacon_blocks_by_{root,range} do not provide optimistic blocks
* handle forkchoice head being pre-merge with block being postmerge
* re-enable blocking head updates on validator duties
* fix is_optimistic_candidate_block per spec; don't crash with nil future
* fix is_optimistic_candidate_block per spec; don't crash with nil future
* mark blocks sans execution payloads valid during head update
* persist LC data across restarts
With the Altair spec `LightClientUpdate` structure taking its final form
it is finally possible to persist LC data across restarts without having
to worry about data migration due to spec changes. A separate `lcdataV1`
database is created in the `caches` subdirectory to hold known LC data.
A full database with default settings (129 periods) uses <15 MB disk.
* extend LC data DB rationale
* wording
* add `isSupportedBySQLite` helper and explicit return
* remove redundant `return`
All message processing is done in the validation callbacks, so there's
no need to trigger data handlers for messages we publish - the
self-publish is async, and therefore has an associated cost
Separate LC initialization options from the main ChainDAGRef options to
allow ChainDAGRef to treat them as opaque and reduce risk for conflicts
when extending those options in the future.
* remove web3 url prompt in launcher script
The interactive prompt for web3 has outlived its utility as we now load
url:s from command line params and config files, preventing the prompt
from correctly detecting when it's needed.
Also, after the merge, a JWT secret will (likely) be needed.
* log notice when web3 url is missing
* fix docs to not mention default that doesn't exist
* fix scripts to properly quote arguments
Merkle proofs tend to have long underlying type definitions, e.g.,
`array[log2trunc(NEXT_SYNC_COMMITTEE_INDEX), Eth2Digest]`. For the
ones used in the LC sync protocol, dedicated types are introduced
to improve readability. Furthermore, the `CachedLightClientBootstrap`
wrapper that solely wrapped a merkle branch is eliminated.
Adds a `--light-client-data-max-periods` option to override the number
of sync committee periods to retain light client data.
Raising it above the default enables archive nodes to serve full data.
Lowering below the default speeds up import times (still no persistence)
This updates `nim-ssz-serialization` to
`3db6cc0f282708aca6c290914488edd832971d61`.
Notable changes:
- Use `uint64` for `GeneralizedIndex`
- Add support for building merkle multiproofs
Combines the LC data configuration options (serve / importMode), the
callbacks (finality / optimistic LC update) as well as the cache storing
light client data, into a new `LightClientDataStore` structure.
Also moves the structure into a light client specific file.
* Initial commit
* Make `events` API spec compliant.
* Add `Eth-Consensus-Version` in responses.
* Bump chronos to get redirect with headers working.
* Add `is_optimistic` field and handling to syncing RestSyncInfo.
* adopt LC REST API with v0 suffix (without proofs)
Adopts the light client data REST API used by Lodestar as defined in
https://github.com/ethereum/beacon-APIs/pull/181 with a v0 suffix.
Requests:
- `/eth/v0/beacon/light_client/bootstrap/{block_root}`
- `/eth/v0/beacon/light_client/updates?start_period={start_period}&count={count}`
- `/eth/v0/beacon/light_client/finality_update`
- `/eth/v0/beacon/light_client/optimistic_update`
HTTP Server-Sent Events (SSE):
- `light_client_finality_update_v0`
- `light_client_optimistic_update_v0`
More work is needed to adopt the proofs endpoint, it is not included.
* initialize event queues
* register event topics
`m.depositsChain.blocks.len` may change during `startEth1Syncing`. If
that happened, an additional check now ensures that `scratchMerkleizer`
was initialized before attempting to use it.
If database access errors are encountered while proccessing LC data,
track the section which was accessed without errors so that the rest
may be attempted to be re-indexed later.
Ensures that all intermediate blocks are reported if a small gap is
encountered when downloading optimistic blocks. Gaps may occur when
a block is missed and still downloading, or when EL processing is slow.
If the gap exceeds 1 epoch, optimistic block stream jumps to latest.
* check for and log gossip broadcast failure
* switch notices to warns; update LC variables regardless
* don't both return a Result and log sending error
* add metrics counter for failed-due-to-no-peers and removed unnecessary async
* don't report failure of sync committee messages
* remove redundant metric
* document metric being incremented
The initial sync committee period follows a different finality rule than
the other ones. Instead of next sync committee finalizing as soon as the
`finalizedHead.slot >= period.start_slot` have to use Altair start slot.
For consistency with other options, use a common prefix for light client
data configuration options.
* `--serve-light-client-data` --> `--light-client-data-serve`
* `--import-light-client-data` --> `--light-client-data-import-mode`
No deprecation of the old identifiers as they were only sparingly used
and all usage can be easily updated without interferance.
Corrects an off-by-1 in the reported sync percentage computation.
New logic is based on `SyncQueue.total` and `SyncQueue.progress`
with `pivot` instead of `sq.startSlot`.
When launched with `--light-client-enable` the latest blocks are fetched
and optimistic candidate blocks are passed to a callback (log for now).
This helps accelerate syncing in the future (optimistic sync).
Adds a `LightClient` instance to the beacon node as preparation to
accelerate syncing in the future (optimistic sync).
- `--light-client-enable` turns on the feature
- `--light-client-trusted-block-root` configures block to start from
If no block root is configured, light client tracks DAG `finalizedHead`.
Introduces a new library for syncing using libp2p based light client
sync protocol, and adds a new `nimbus_light_client` executable that uses
this library for syncing. The new executable emits log messages when
new beacon block headers are received, and is integrated into testing.
* SSZ `[]` -> `mitem`
* `[]` -> `item`
immutable access via mutable instance cannot rely on template
overloading, and `[]` cannot be a `func` because of special seq handling
in compiler.
* remove deprecated JSON-RPC server
* keep the command-line options around as no-ops, temporarily
* service -> server; JSON-RPC is still used elsewhere
* document static vs dynamic range checking requirements
* add `vindices` iterator to iterate over valid validator indices in a
state
* clean up spec comments in general
* fixup
Co-authored-by: tersec <tersec@users.noreply.github.com>
Incorporates the latest changes to the light client sync protocol based
on Devconnect AMS feedback. Note that this breaks compatibility with the
previous prototype, due to changes to data structures and endpoints.
See https://github.com/ethereum/consensus-specs/pull/2802
Since we were not verifying BLS signature in blocks that we produce,
we were failing to notice that some deposits need to be ignored (due
to having an invalid signature). Processing these deposits resulted
in a different ending state after the state transition which caused
our blocks to be rejected by the network.
* Some Web3Signer versions insist replying with text/plain messages
* When reading blocks, the Web3Signer uses upper-case fork identifiers
instead of lower-case identifies like the Beacon API.
Follows up on https://github.com/status-im/nimbus-eth2/pull/3461 which
ensured that repeated `beaconBlocksByRange` requests get shrinked to
account for potential out-of-band advancements to `safeSlot`, with
similar logic for the initial request.
Other changes:
* logtrace can now verify sync committee messages and contributions
* Many unnecessary use of pairs() have been removed for consistency
* Map 40x BN response codes to BeaconNodeStatus.Incompatible in the VC
Other fixes:
* Fix bit rot in the `make prater-dev-deposit` target.
* Correct content-type in the responses of the Nimbus signing node
* Invalid JSON payload was being sent in the web3signer requests
* era file verification
Implement and document era file verification
* era file states now come with block applied for easier verification
* clarify conflicting version handling
* document verification requirements
* remove count from name, use start-era, end-root to discover range
* remove obsolete todo
* abstract out block root loading
Updated outdated presets / configs / REST config to v1.1.10 specs.
- `TERMINAL_BLOCK_HASH_ACTIVATION_EPOCH` and `PROPOSER_SCORE_BOOST` are
not yet used in `eth2-networks`, added configurability as TODOs.
- `MIN_ANCHOR_POW_BLOCK_DIFFICULTY` is no longer needed, put on ignore
list as some Altair devnets still reference it.
* use MAX_CHUNK_SIZE_BELLATRIX for signed Bellatrix blocks
* Update beacon_chain/networking/eth2_network.nim
Co-authored-by: Etan Kissling <etan@status.im>
* localPassC to localPassc
* check against maxChunkSize rather than constant
Co-authored-by: Etan Kissling <etan@status.im>
This PR makes the necessary adjustments to deal with the revamped snappy
API.
In practical terms for nimbus-eth2, there are performance increases to
gossip processing, database reading and writing as well as era file
processing. Exporting `.era` files for example, a snappy-heavy
operation, almost halves in total processing time:
Pre:
```
Average, StdDev, Min, Max, Samples, Test
39.088, 8.735, 23.619, 53.301, 50, tState
237.079, 46.692, 165.620, 355.481, 49, tBlocks
```
Post:
```
All time are ms
Average, StdDev, Min, Max, Samples, Test
25.350, 5.303, 15.351, 41.856, 50, tState
141.238, 24.164, 99.990, 199.329, 49, tBlocks
```
* Add `NoMonitor` flag to stop SyncManager from monitoring sync situation.
* Remove `toleranceValue` and `PeerScoreHeadTooNew`.
Co-authored-by: Etan Kissling <etan@status.im>
* `gnosis-chain` -> `gnosis`
Use same name as LH/Teku throughout
* fixes#3504
* fixes large stack temporary that can cause crashes during genesis
detection
Some upstream repos still need fixes, but this gets us close enough that
style hints can be enabled by default.
In general, "canonical" spellings are preferred even if they violate
nep-1 - this applies in particular to spec-related stuff like
`genesis_validators_root` which appears throughout the codebase.
Validator monitoring improves logging by giving more specific monitoring
information, and can now be seen as complete.
Previously, logging has focused on "Attestation sent" messages which
carry little informational value when things go wrong, and are overly
aggressive when everything works as expected (sending attestations is
the norm).
* lower "Attestation sent" log to `INFO`
* mark 1.7.0 as the start of the validator monitor feature - previous
versions had significant bugs in totals mode
The `pyrmont` testnet has been discontinued.
For experiments, it's still possible to run pyrmont nodes by passing a
genesis/config, but this PR removes the bundled `--network:pyrmont`
option.
* update docs
* remove empty docs
* remove obsolete `eth2-stats` page
`.era` files and Req/Resp protocols use framed formats - aligning the
database with these makes for less recompression work overall as gossip
is sent only once while req/resp repeats (potentially) - this also
allows efficient pruning-to-era where snappy-recompression is the major
cycle thief.
* harden validator API against pre-finalized slot requests
* check `syncHorizon` when responding to validator api requests too far
from `head`
* limit state-id based requests to one epoch ahead of `head`
* put historic data bounds on block/attestation/etc validator production API, preventing them from being used with already-finalized slots
* add validator block smoke tests
* make rest test create a new genesis with the tests running roughly in
the first epoch to allow testing a few more boundary conditions
* era: load blocks and states
Era files contain finalized history and can be thought of as an
alternative source for block and state data that allows clients to avoid
syncing this information from the P2P network - the P2P network is then
used to "top up" the client with the most recent data. They can be
freely shared in the community via whatever means (http, torrent, etc)
and serve as a permanent cold store of consensus data (and, after the
merge, execution data) for history buffs and bean counters alike.
This PR gently introduces support for loading blocks and states in two
cases: block requests from rest/p2p and frontfilling when doing
checkpoint sync.
The era files are used as a secondary source if the information is not
found in the database - compared to the database, there are a few key
differences:
* the database stores the block indexed by block root while the era file
indexes by slot - the former is used only in rest, while the latter is
used both by p2p and rest.
* when loading blocks from era files, the root is no longer trivially
available - if it is needed, it must either be computed (slow) or cached
(messy) - the good news is that for p2p requests, it is not needed
* in era files, "framed" snappy encoding is used while in the database
we store unframed snappy - for p2p2 requests, the latter requires
recompression while the former could avoid it
* front-filling is the process of using era files to replace backfilling
- in theory this front-filling could happen from any block and
front-fills with gaps could also be entertained, but our backfilling
algorithm cannot take advantage of this because there's no (simple) way
to tell it to "skip" a range.
* front-filling, as implemented, is a bit slow (10s to load mainnet): we
load the full BeaconState for every era to grab the roots of the blocks
- it would be better to partially load the state - as such, it would
also be good to be able to partially decompress snappy blobs
* lookups from REST via root are served by first looking up a block
summary in the database, then using the slot to load the block data from
the era file - however, there needs to be an option to create the
summary table from era files to fully support historical queries
To test this, `ncli_db` has an era file exporter: the files it creates
should be placed in an `era` folder next to `db` in the data directory.
What's interesting in particular about this setup is that `db` remains
as the source of truth for security purposes - it stores the latest
synced head root which in turn determines where a node "starts" its
consensus participation - the era directory however can be freely shared
between nodes / people without any (significant) security implications,
assuming the era files are consistent / not broken.
There's lots of future improvements to be had:
* we can drop the in-memory `BlockRef` index almost entirely - at this
point, resident memory usage of Nimbus should drop to a cool 500-600 mb
* we could serve era files via REST trivially: this would drop backfill
times to whatever time it takes to download the files - unlike the
current implementation that downloads block by block, downloading an era
at a time almost entirely cuts out request overhead
* we can "reasonably" recreate detailed state history from almost any
point in time, turning an O(slot) process into O(1) effectively - we'll
still need caches and indices to do this with sufficient efficiency for
the rest api, but at least it cuts the whole process down to minutes
instead of hours, for arbitrary points in time
* CI: ignore failures with Nim-1.6 (temporary)
* test fixes
Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
This extends the `--serve-light-client-data` launch option to serve
locally collected light client data via libp2p.
Backfill of historic best `LightClientUpdate` is not yet implemented.
See https://github.com/ethereum/consensus-specs/pull/2802
During operation as a light client, the chain DAG is not available.
As a preparation, the beacon node initialization logic is divided into
parts depending on the presence of the chain DAG, and parts that are
always available (including a future light client mode).
This is a pure code move without semantic changes.
Gracefully handles the new failure modes recently introduced to the DAG
as part of https://github.com/status-im/nimbus-eth2/pull/3513
Data that is deemed to exist but fails to load leads to an error log to
avoid suppressing logic errors accidentally. In `verifyFinalization`
mode, the assertions remain active.
When eliminating orphaned forks, light client data about blocks was also
deleted when the orphaned fork was referring to a state several slots
after the block. Linking light client data pruning with block deletion
instead of state deletion fixes this problem. Light client data always
refers to blocks and their immediate post-state.
When transitioning from light client to full node the chain DAG will be
loaded separately from the rest of the beacon node initialization.
Extracting chain DAG loading to a separate function will allow reusing
a lot of the existing code. This code move doesn't change semantics.
ref loop would stop one block early in this case - trying to load
everything in one loop ends up being pretty confusing..
* simplify finalizedBlocks topup by splitting it from the head loop /
query
When doing checkpoint sync, collecting light client data of known blocks
and states incorrectly assumes that `finalized_checkpoint` information
is also known. Hardens collection to only collect finalized checkpoint
data after `dag.computeEarliestLightClientSlot`.
This file is not actually used / useful - should metadata persistence
support be added in the future, it needs to be done with a new file such
that downgrades, that have the TODO logic unimplemented, don't break.
Witout this, we end up with a massive .wal file that needs to be
checkpointed on first startup (which takes a few minutes) - it's much
more efficient to do smaller checkpoints, it turns out.
Recently, block processing times have been going up as the network grows
making early attestation riskier. Since blocks are big and attestations
are small (though numerous and therefore bandwidth-intense), it seems
better to wait a little bit longer after receiving a block, before we
publish the attestation.
Adds `LightClientProcessor` as the pendant to `BlockProcessor` while
operating in light client mode. Note that a similar mechanism based on
async futures is used for interoperability with existing infrastructure,
despite light client object validation being done synchronously.
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 :)
The pre-release light client sync protocol defines additional Req/Resp
messages to be made available when `--serve-light-client-data` is set.
This patch extends the `{.libp2pProtocol.}` pragma with an optional
parameter to tag such light client sync protocol specific messages.
The corresponding protocols are only selectively registered with libp2p.
The spec implicitly talks about the slot of a block in several places,
and keeping it readily available is useful in a number of context -
might as well put this implicitly refereneced helper in the spec code
directly
One more step on the journey to reduce `BlockRef` usage across the
codebase - this one gets rid of `StateData` whose job was to keep track
of which block was last assigned to a state - these duties have now been
taken over by `latest_block_root`, a fairly recent addition that
computes this block root from state data (at a small cost that should be
insignificant)
99% mechanical change.
When a `beaconBlocksByRange` response advances the `safeSlot`, but later
has errors, the sync queue keeps repeating that same request until it is
fulfilled without errors. Data up through `safeSlot` is considered to be
immutable, i.e., finalized, so re-requesting that data is not useful.
By advancing the sync progress in that scenario, those redundant query
portions can be avoided. Note, the finalized block _itself_ is always
requested, even in the initial request. This behaviour is kept same.
* fewer deps on `BlockRef` traversal in anticipation of pruning
* allows identifying EpochRef:s by their shuffling as a first step of
* tighten error handling around missing blocks
using the zero hash for signalling "missing block" is fragile and easy
to miss - with checkpoint sync now, and pruning in the future, missing
blocks become "normal".
When syncing as a light client, different behaviour is needed to handle
the various ways how errors may occur. The existing logic for blocks can
also be applied to light client objects:
- `Invalid`: Malformed object that is clearly an error by its producer.
- `MissingParent`: More data is needed to decide applicability.
- `UnviableFork`: Object may be valid but will never apply on this fork.
- `Duplicate`: No errors were encountered but the object was not useful.
Light clients require full nodes to serve additional data so that they
can stay in sync with the network. This patch adds a new launch option
`--import-light-client-data` to configure what data to make available.
For now, data is only kept in memory; it is not persisted at this time.
Note that data is only locally collected, a separate patch is needed to
actually make it availble over the network. `--serve-light-client-data`
will be used for serving data, but is not functional yet outside tests.
When performing trusted node sync, historical access is limited to
states after the checkpoint.
Reindexing restores full historical access by replaying historical
blocks against the state and storing snapshots in the database.
The process can be initiated or resumed at any point in time.
The `p2p-interface.md` spec defines a `ResourceUnavailable` error to
return in situations where data that exists on the network is locally
unavailable, e.g., when a block within `MIN_EPOCHS_FOR_BLOCK_REQUESTS`
is requested by `BeaconBlocksByRange` but cannot be provided. This patch
adds support for that additional error code.
This adopts the spec sections of the pre-release proposal of the libp2p
based light client sync protocol, and also adds a test runner for the
new accompanying tests. While the release version of the light client
sync protocol contains conflicting definitions, it is currently unused,
and the code specific to the pre-release proposal is marked as such.
See https://github.com/ethereum/consensus-specs/pull/2802
`BlockId` is a type that bundles a block root with its slot number.
The type can be useful as key in tables that deal with non-finalized
blocks (not uniquely identified by slot) and also support pruning
(drop data about older blocks by slot). Instead of creating a custom
type for those use cases, this patch suggests implementing `hash` for
`BlockId` to re-use the existing type.
* --stop-at-synced-epoch
This allows benchmarking the initial sync (only forward sync, 1s error
margin). Might be useful in CI, with a timeout, as a sanity check.
The spec does not provide code for validating the `fork_version` field
of `LightClientUpdate`. However, we can use our own logic for additional
validation of that field. The spec's python test suite sets up states
that do not follow the fork schedule (e.g., that use Altair fork version
before Altair fork epoch), which complicates upstreaming this as code.
* Refactor and optimize logs.
* Introduce shortLog(SyncRequest).
* Address review comment.
* make sync queue logs more consistent
Adds a few minor logging improvements:
- Fixes a typo (`was happened` -> `has happened`)
- Avoids passing `reset_slot` argument to log statement multiple times
- Uses same `rewind_to_slot` label when logging in both sync directions
- Consistent rewind point logging
Co-authored-by: cheatfate <eugene.kabanov@status.im>
Uses consistent formatting in `light_client_sync.nim`, always refers to
fork-dependent light client objects in full qualified notation, moves
`get_safety_threshold` helper function to same location as in the spec.
Can't apply a phase0 block to a later phase state and vice versa.
Since instantiation has been a topic, pre/post c file size:
```
424K @mspec@sstate_transition.nim.c
892K @mspec@sstate_transition_block.nim.c
```
```
288K @mspec@sstate_transition.nim.c
880K @mspec@sstate_transition_block.nim.c
```
Updates the spec references for `GeneralizedIndex` constants used by the
light client sync protocol, and adds a short explanation how they are
derived and which SSZ fields they refer to.
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
https://github.com/ethereum/consensus-specs/pull/2225 removed an ignore
rule that would filter out duplicate aggregates from gossip publishing -
however, this causes increased bandwidth and CPU usage as discussed in
https://github.com/ethereum/consensus-specs/issues/2183 - the intent is
to revert the removal and reinstate the rule.
This PR implements ignore filtering which cuts down on CPU usage (fewer
aggregates to validate) and bandwidth usage (less fanout of duplicates)
- as #2225 points out, this may lead to a small increase in IHAVE
messages.
* Support for Gnosis Chain
`make gnosis-chain-build` will build the Nimbus gnosis chain binary,
stored in `build/nimbus_beacon_node_for_gnosis_chain`.
`make gnosis-chain` will connect to the network.
Other changes:
* Restore compilation with -d:has_genesis_detection
* Removed Makefile target related to testnet0 and testnet1
* Added more debug logging for failed peer handshakes
* Report misconfigured builds which try to embed network metadata
that is incompatible with the currently selected const preset.
* Don't bundle network metadata in minimal builds, as they are not compatible
To calculate the deltas correctly, the `process_inactivity_updates` function
must be called before the rewards and penalties processing code in order to
update the `inactivity_scores` field in the state. This would have required
duplicating more logic from the spec in the ncli modules, so I've decided to
pay the price of introducing a run-time copy of the state at each epoch which
eliminates the need to duplicate logic (both for this fix and the previous one).
Other changes:
* Fixes for the read-only mode of the `BeaconChainDb`
* Fix an uint64 underflow in the debug output procedure for printing
balance deltas
* Allow Bellatrix states in the reward computation helpers
Streamline lookup with Forky and BeaconBlockFork (then we can do the
same for era)
We use type to avoid conditionals, as fork is often already known at a
"higher" level.
* load blockid before loading block by root - this is needed to map root
to slot and will eventually be done via block summary table for "old"
blocks
Co-authored-by: tersec <tersec@users.noreply.github.com>
Update several `ncli_db` commands to run in readOnly mode, allowing them
to be used with a running instance - in particular era export.
* export all eras by default
* skip already-exported eras
* clean up / document init
* drop `immutable_validators` data (pre-altair)
* document versions where data is first added
* avoid needlessly loading genesis block data on startup
* add a few more internal database consistency checks
* remove duplicate state root lookup on state load
* comment
When initializing backfill sync, the implementation intends to start at
the first unknown slot (`1` before tail). However, an incorrect variable
is passed, and backfill sync actually starts at the tail slot instead.
This patch corrects this by passing the intended variable. The problem
was introduced with the original backfill implementation at #3263.
The added options work in opt-in fashion. If they are not specified,
the server will respond to all requests as if the CORS specification
doesn't exist. This will result in errors in CORS-enabled clients.
Please note that future versions may support more than one allowed
origin. The option names will stay the same, but the user will be
able to repeat them on the command line (similar to other options
such as --web3-url).
To be documented in the guide in a separate PR.
The `SyncManager` has a leftover optional `sleepTime` parameter in
its constructor that used to configure the sync loop polling rate.
This parameter was replaced with a constant in #1602 and is no longer
functional. This patch removes the `sleepTime` leftovers.
#3304 introduced a regression to the sync status string displayed in the
status bar; during the main forward sync, the current slot is no longer
reported and always displays as `0`. This patch corrects the computation
to accurately report the current slot once more.
The `SyncManager` has a leftover optional `maxStatusAge` parameter in
its constructor that used to configure the libp2p `Status` polling rate.
This parameter was replaced with a constant in #1827 and is no longer
functional. This patch removes the `maxStatusAge` leftovers.
With these changes, we can backfill about 400-500 slots/sec, which means
a full backfill of mainnet takes about 2-3h.
However, the CPU is not saturated - neither in server nor in client
meaning that somewhere, there's an artificial inefficiency in the
communication - 16 parallel downloads *should* saturate the CPU.
One plasible cause would be "too many async event loop iterations" per
block request, which would introduce multiple "sleep-like" delays along
the way.
I can push the speed up to 800 slots/sec by increasing parallel
downloads even further, but going after the root cause of the slowness
would be better.
* avoid some unnecessary block copies
* double parallel requests
When node is restarted before backfill has started but after some blocks
have finalized with forward sync, we would not start the backfill.
* also clean up one last `SomeSome`
* Initial commit.
* Fix current test suite.
* Fix keymanager api test.
* Fix wss_sim.
* Add more keystore_management tests.
* Recover deleted isEmptyDir().
* Add `HttpHostUri` distinct type.
Move keymanager calls away from rest_beacon_calls to rest_keymanager_calls.
Add REST serialization of RemoteKeystore and Keystore object.
Add tests for Remote Keystore management API.
Add tests for Keystore management API (Add keystore).
Fix serialzation issues.
* Fix test to use HttpHostUri instead of Uri.
* Add links to specification in comments.
* Remove debugging echoes.
* harden and speed up block sync
The `GetBlockBy*` server implementation currently reads SSZ bytes from
database, deserializes them into a Nim object then serializes them right
back to SSZ - here, we eliminate the deser/ser steps and send the bytes
straight to the network. Unfortunately, the snappy recoding must still
be done because of differences in framing.
Also, the quota system makes one giant request for quota right before
sending all blocks - this means that a 1024 block request will be
"paused" for a long time, then all blocks will be sent at once causing a
spike in database reads which potentially will see the reading client
time out before any block is sent.
Finally, on the reading side we make several copies of blocks as they
travel through various queues - this was not noticeable before but
becomes a problem in two cases: bellatrix blocks are up to 10mb (instead
of .. 30-40kb) and when backfilling, we process a lot more of them a lot
faster.
* fix status comparisons for nodes syncing from genesis (#3327 was a bit
too hard)
* don't hit database at all for post-altair slots in GetBlock v1
requests
* deactivate Doppelganger Protection during genesis
* also don't actually flag supposed-doppelgangers (because they're before broadcastStartEpoch) on GENESIS_SLOT start
* update action tracker on dependent-root-changing reorg (instead of
epoch change)
* don't try to log duties while syncing - we're not tracking actions yet
* fix slot used for doppelganger loss detection
These use a separate flow, and were previously only registered from the
network
* don't log successes in totals mode (TMI)
* remove `attestation-sent` event which is unused
* Fix a resource leak introduced in https://github.com/status-im/nimbus-eth2/pull/3279
* Don't restart the Eth1 syncing proggress from scratch in case of
monitor failures during Eth2 syncing.
* Switch to the primary operator as soon as it is back online.
* Log the web3 credentials in fewer places
Other changes:
The 'web3 test' command has been enhanced to obtain and print more
data regarding the selected provider.
- Request metadata_v2 (altair) by default instead of the v1
- Change the metadata pinger to a 3 failure-then-kick, instead of being time based
- Update kicker scorer to take into account topics which we're not subscribed to, to be sure that we will be able to publish correctly
- Add some metrics to give "fanout" health (in the same spirit of mesh health)
Make `validator exit command` work both with `JSON-RPC` and `REST` APIs
Fix problem with specifying rest-url using `localhost`
Change back exit error messages in `state_transition_block`
The current counters set gauges etc to the value of the _last_ validator
to be processed - as the name of the feature implies, we should be using
sums instead.
* fix missing beacon state metrics on startup, pre-first-head-selection
* fix epoch metrics not being updated on cross-epoch reorg
* Store finalized block roots in database (3s startup)
When the chain has finalized a checkpoint, the history from that point
onwards becomes linear - this is exploited in `.era` files to allow
constant-time by-slot lookups.
In the database, we can do the same by storing finalized block roots in
a simple sparse table indexed by slot, bringing the two representations
closer to each other in terms of conceptual layout and performance.
Doing so has a number of interesting effects:
* mainnet startup time is improved 3-5x (3s on my laptop)
* the _first_ startup might take slightly longer as the new index is
being built - ~10s on the same laptop
* we no longer rely on the beacon block summaries to load the full dag -
this is a lot faster because we no longer have to look up each block by
parent root
* a collateral benefit is that we no longer need to load the full
summaries table into memory - we get the RSS benefits of #3164 without
the CPU hit.
Other random stuff:
* simplify forky block generics
* fix withManyWrites multiple evaluation
* fix validator key cache not being updated properly in chaindag
read-only mode
* drop pre-altair summaries from `kvstore`
* recreate missing summaries from altair+ blocks as well (in case
database has lost some to an involuntary restart)
* print database startup timings in chaindag load log
* avoid allocating superfluos state at startup
* use a recursive sql query to load the summaries of the unfinalized
blocks
Additional sanity checking of the status message exchanged during a
fresh connection:
* check that head and finalized make sense, slot-wise
* verify that finalized root lies on the canonical chain, when possible
* re-check these things for every status message during sync
* Harden handling of unviable forks
In our current handling of unviable forks, we allow peers to send us
blocks that come from a different fork - this is not necessarily an
error as it can happen naturally, but it does open up the client to a
case where the same unviable fork keeps getting requested - rather than
allowing this to happen, we'll now give these peers a small negative
score - if it keeps happening, we'll disconnect them.
* keep track of unviable forks in quarantine, to avoid filling it with
known junk
* collect peer scores in single module
* descore peers when they send unviable blocks during sync
* don't give score for duplicate blocks
* increase quarantine size to a level that allows finality to happen
under optimal conditions - this helps avoid downloading the same blocks
over and over in case of an unviable fork
* increase initial score for new peers to make room for one more failure
before disconnection
* log and score invalid/unviable blocks in requestmanager too
* avoid ChainDAG dependency in quarantine
* reject gossip blocks with unviable parent
* continue processing unviable sync blocks in order to build unviable
dag
* docs
* Update beacon_chain/consensus_object_pools/block_pools_types.nim
* add unviable queue test
* 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
* initial support for minification and new interchange tests. Removal of v1 and v1 migration.
* Synthetic attestations: SQLite3 requires one statement/query per prepared statement
* Fix DB import interrupted if no attestation was found
* Skip test relying on undocumented test behavior (https://github.com/eth-clients/slashing-protection-interchange-tests/pull/12#issuecomment-1011158701)
* Skip test relying on unclear minification behavior:
creating an invalid minified attestation with source > target or setting target = max(source, target)
* remove DB v1 and update submodule
* Apply suggestions from code review
Co-authored-by: Jacek Sieka <jacek@status.im>
Co-authored-by: Jacek Sieka <jacek@status.im>
Backfilling is the process of downloading historical blocks via P2P that
are required to fulfill `GetBlocksByRange` duties - this happens during
both trusted node and finalized checkpoint syncs.
In particular, backfilling happens after syncing to head, such that
attestation work can start as soon as possible.
* Fix SyncQueue initialization procedure.
Remove usage of `awaitne`.
Add cancellation support.
Remove unneeded `sleepAsync()` if peer's head is older than needed.
Add `direction` field to all logs.
Fix syncmanager wedge issue.
Add proper resource cleaning procedure on backward sync finish.
Co-authored-by: cheatfate <eugene.kabanov@status.im>
The new format is based on compressed CSV files in two channels:
* Detailed per-epoch data
* Aggregated "daily" summaries
The use of append-only CSV file speeds up significantly the epoch
processing speed during data generation. The use of compression
results in smaller storage requirements overall. The use of the
aggregated files has a very minor cost in both CPU and storage,
but leads to near interactive speed for report generation.
Other changes:
- Implemented support for graceful shut downs to avoid corrupting
the saved files.
- Fixed a memory leak caused by lacking `StateCache` clean up on each
iteration.
- Addressed review comments
- Moved the rewards and penalties calculation code in a separate module
Required invasive changes to existing modules:
- The `data` field of the `KeyedBlockRef` type is made public to be used
by the validator rewards monitor's Chain DAG update procedure.
- The `getForkedBlock` procedure from the `blockchain_dag.nim` module
is made public to be used by the validator rewards monitor's Chain DAG
update procedure.
This is an alternative take on https://github.com/status-im/nimbus-eth2/pull/3107
that aims for more minimal interventions in the spec modules at the expense of
duplicating more of the spec logic in ncli_db.
Using the wrong type here causes requests to fail due to the overly
zealous parameter validation - the failure is harmless in the current
duty subscription model, but would have caused more serious failures
down the line.
* Trusted node sync
Trusted node sync, aka checkpoint sync, allows syncing tyhe chain from a
trusted node instead of relying on a full sync from genesis.
Features include:
* sync from any slot, including the latest finalized slot
* backfill blocks either from the REST api (default) or p2p (#3263)
Future improvements:
* top up blocks between head in database and some other node - this
makes for an efficient backup tool
* recreate historical state to enable historical queries
* fixes
* load genesis from network metadata
* check checkpoint block root against state
* fix invalid block root in rest json decoding
* odds and ends
* retry looking for epoch-boundary checkpoint blocks
When syncing, we show how much of the sync has completed - with
checkpoint sync, the syncing does not always go from slot 0 to head, but
rather can start in the middle.
To show a consistent `%` between restarts, we introduce the concept of a
pivot point, such that if I sync 10% of the chain, then restart the
client, it picks up at 10% (instead of counting from 0).
What it looks like:
```
INF ... sync="01d12h41m (15.96%) 13.5158slots/s (QDDQDDQQDP:339018)" ...
```
Time in the beacon chain is expressed relative to the genesis time -
this PR creates a `beacon_time` module that collects helpers and
utilities for dealing the time units - the new module does not deal with
actual wall time (that's remains in `beacon_clock`).
Collecting the time related stuff in one place makes it easier to find,
avoids some circular imports and allows more easily identifying the code
actually needs wall time to operate.
* move genesis-time-related functionality into `spec/beacon_time`
* avoid using `chronos.Duration` for time differences - it does not
support negative values (such as when something happens earlier than it
should)
* saturate conversions between `FAR_FUTURE_XXX`, so as to avoid
overflows
* fix delay reporting in validator client so it uses the expected
deadline of the slot, not "closest wall slot"
* simplify looping over the slots of an epoch
* `compute_start_slot_at_epoch` -> `start_slot`
* `compute_epoch_at_slot` -> `epoch`
A follow-up PR will (likely) introduce saturating arithmetic for the
time units - this is merely code moves, renames and fixing of small
bugs.
* Harden CommitteeIndex, SubnetId, SyncSubcommitteeIndex
Harden the use of `CommitteeIndex` et al to prevent future issues by
using a distinct type, then validating before use in several cases -
datatypes in spec are kept simple though so that invalid data still can
be read.
* fix invalid epoch used in REST
`/eth/v1/beacon/states/{state_id}/committees` committee length (could
return invalid data)
* normalize some variable names
* normalize committee index loops
* fix `RestAttesterDuty` to use `uint64` for `validator_committee_index`
* validate `CommitteeIndex` on ingress in REST API
* update rest rules with stricter parsing
* better REST serializers
* save lots of memory by not using `zip` ...at least a few bytes!
* REST cleanups
* reject out-of-range committee requests
* print all hex values as lower-case
* allow requesting state information by head state root
* turn `DomainType` into array (follow spec)
* `uint_to_bytesXX` -> `uint_to_bytes` (follow spec)
* fix wrong dependent root in `/eth/v1/validator/duties/proposer/`
* update documentation - `--subscribe-all-subnets` is no longer needed
when using the REST interface with validator clients
* more fixes
* common helpers for dependent block
* remove test rules obsoleted by more strict epoch tests
* fix trailing commas
* Update docs/the_nimbus_book/src/rest-api.md
* Update docs/the_nimbus_book/src/rest-api.md
Co-authored-by: sacha <sacha@status.im>
Eugene Kabanov
Jacek Sieka
Etan Kissling
tersec
Michael Sproul
Zahary Karadjov
Miran
Tanguy
Taneli Hukkinen
zah
Ștefan Talpalaru
Mamy Ratsimbazafy
EmilIvanichkovv