why:
Single mode here means there is only such (single mode) instance
activated but multi mode instances for other peers are allowed.
Erroneously, multi mode instances were held back waiting while some
single mode instance was running which reduced the number of parallel
download peers.
* Update log ticker, using time interval rather than ticker count
why:
Counting and logging ticker occurrences is inherently imprecise. So
time intervals are used.
* Use separate storage tables for snap sync data
* Left boundary proof update
why:
Was not properly implemented, yet.
* Capture pivot in peer worker (aka buddy) tasks
why:
The pivot environment is linked to the `buddy` descriptor. While
there is a task switch, the pivot may change. So it is passed on as
function argument `env` rather than retrieved from the buddy at
the start of a sub-function.
* Split queues `fetchStorage` into `fetchStorageFull` and `fetchStoragePart`
* Remove obsolete account range returned from `GetAccountRange` message
why:
Handler returned the wrong right value of the range. This range was
for convenience, only.
* Prioritise storage slots if the queue becomes large
why:
Currently, accounts processing is prioritised up until all accounts
are downloaded. The new prioritisation has two thresholds for
+ start processing storage slots with a new worker
+ stop account processing and switch to storage processing
also:
Provide api for `SnapTodoRanges` pair of range sets in `worker_desc.nim`
* Generalise left boundary proof for accounts or storage slots.
why:
Detailed explanation how this works is documented with
`snapdb_accounts.importAccounts()`.
Instead of enforcing a left boundary proof (which is still the default),
the importer functions return a list of `holes` (aka node paths) found in
the argument ranges of leaf nodes. This in turn is used by the book
keeping software for data download.
* Forgot to pass on variable in function wrapper
also:
+ Start healing not before 99% accounts covered (previously 95%)
+ Logging updated/prettified
* Update docu and logging
* Extracted and updated constants from `worker_desc` into separate file
* Update and re-calibrate communication error handling
* Allow simplified pivot negotiation
why:
This feature allows to turn off pivot negotiation so that peers agree
on a a pivot header.
For snap sync with fast changing pivots this only throttles the sync
process. The finally downloaded DB snapshot is typically a merged
version of different pivot states augmented by a healing process.
* Re-model worker queues for accounts download & healing
why:
Currently there is only one data fetch per download or healing task.
This task is then repeated by the scheduler after a short time. In
many cases, this short time seems enough for some peers to decide to
terminate connection.
* Update main task batch `runMulti()`
details:
The function `runMulti()` is activated in quasi-parallel mode by the
scheduler. This function calls the download, healing and fast-sync
functions.
While in debug mode, after each set of jobs run by this function the
database is analysed (by the `snapdb_check` module) and the result
printed.
* Update logging
* Fix node hash associated with partial path for missing nodes
why:
Healing uses the partial paths for fetching nodes from the network. The
node hash (or key) is used to verify the node data retrieved.
The trie inspector function returned the parent hash instead of the node hash
with the partial path when a missing node was detected. So all nodes
for healing were rejected.
* Must not modify sequence while looping over it
* Re-arrange fetching storage slots in batch module
why;
Previously, fetching partial slot ranges first has a chance of
terminating the worker peer 9due to network error) while there were
many inheritable storage slots on the queue.
Now, inheritance is checked first, then full slot ranges and finally
partial ranges.
* Update logging
* Bundled node information for healing into single object `NodeSpecs`
why:
Previously, partial paths and node keys were kept in separate variables.
This approach was error prone due to copying/reassembling function
argument objects.
As all partial paths, keys, and node data types are more or less handled
as `Blob`s over the network (using Eth/6x, or Snap/1) it makes sense to
hold these `Blob`s as named field in a single object (even if not all
fields are active for the current purpose.)
* For good housekeeping, using `NodeKey` type only for account keys
why:
previously, a mixture of `NodeKey` and `Hash256` was used. Now, only
state or storage root keys use the `Hash256` type.
* Always accept latest pivot (and not a slightly older one)
why;
For testing it was tried to use a slightly older pivot state root than
available. Some anecdotal tests seemed to suggest an advantage so that
more peers are willing to serve on that older pivot. But this could not
be confirmed in subsequent tests (still anecdotal, though.)
As a side note, the distance of the latest pivot to its predecessor is
at least 128 (or whatever the constant `minPivotBlockDistance` is
assigned to.)
* Reshuffle name components for some file and function names
why:
Clarifies purpose:
"storages" becomes: "storage slots"
"store" becomes: "range fetch"
* Stash away currently unused modules in sub-folder named "notused"
* Multiple storage batches at a time
why:
Previously only some small portion was processed at a time so the peer
might have gone when the process was resumed at a later time
* Renamed some field of snap/1 protocol response object
why:
Documented as `slots` is in reality a per-account list of slot lists. So
the new name `slotLists` better reflects the nature of the beast.
* Some minor healing re-arrangements for storage slot tries
why;
Resolving all complete inherited slots tries first in sync mode keeps
the worker queues smaller which improves logging.
* Prettify logging, comments update etc.
* Re-model persistent database access
why:
Storage slots healing just run on the wrong sub-trie (i.e. the wrong
key mapping). So get/put and bulk functions now use the definitions
in `snapdb_desc` (earlier there were some shortcuts for `get()`.)
* Fixes: missing return code, typo, redundant imports etc.
* Remove obsolete debugging directives from `worker_desc` module
* Correct failing unit tests for storage slots trie inspection
why:
Some pathological cases for the extended tests do not produce any
hexary trie data. This is rightly detected by the trie inspection
and the result checks needed to adjusted.
* For snap sync, publish `EthWireRef` in sync descriptor
why:
currently used for noise control
* Detect and reuse existing storage slots
* Provide healing module for storage slots
* Update statistic ticker (adding range factor for unprocessed storage)
* Complete mere function for work item ranges
why:
Merging interval into existing partial item was missing
* Show av storage queue lengths in ticker
detail;
Previous attempt shows average completeness which did not tell much
* Correct the meaning of the storage counter (per pivot)
detail:
Is the # accounts that have a storage saved
* Rename `LeafRange` => `NodeTagRange`
* Replacing storage slot partition point by interval
why:
The partition point only allows to describe slots `[point,high(Uint256)]`
for fetching interval slot ranges. This has been generalised for any
interval.
* Replacing `SnapAccountRanges` by `SnapTrieRangeBatch`
why:
Generalised healing status for accounts, and later for storage slots.
* Improve accounts healing loop
* Split `snap_db` into accounts and storage modules
why:
It is cleaner to have separate session descriptors for accounts and
storage slots (based on a common base descriptor.)
Also, persistent storage handling might be changed in future which
requires the storage slot implementation disentangled from the accounts
handling.
* Re-model worker queues for storage slots
why:
There is a dynamic list of storage sub-tries, each one has to be
treated similar to the accounts database. This applied to slot
interval downloads as well as to healing
* Compress some return value report lists for snapdb methods
why:
No need to report all handling details for work items that are filteres
out and discarded, anyway.
* Remove inner loop frame from healing function
why:
The healing function runs as a loop body already.
* Split fetch accounts into sub-modules
details:
There will be separated modules for accounts snapshot, storage snapshot,
and healing for either.
* Allow to rebase pivot before negotiated header
why:
Peers seem to have not too many snapshots available. By setting back the
pivot block header slightly, the chances might be higher to find more
peers to serve this pivot. Experiment on mainnet showed that setting back
too much (tested with 1024), the chances to find matching snapshot peers
seem to decrease.
* Add accounts healing
* Update variable/field naming in `worker_desc` for readability
* Handle leaf nodes in accounts healing
why:
There is no need to fetch accounts when they had been added by the
healing process. On the flip side, these accounts must be checked for
storage data and the batch queue updated, accordingly.
* Reorganising accounts hash ranges batch queue
why:
The aim is to formally cover as many accounts as possible for different
pivot state root environments. Formerly, this was tried by starting the
accounts batch queue at a random value for each pivot (and wrapping
around.)
Now, each pivot environment starts with an interval set mutually
disjunct from any interval set retrieved with other pivot state roots.
also:
Stop fishing for more pivots in `worker` if 100% download is reached
* Reorganise/update accounts healing
why:
Error handling was wrong and the (math. complexity of) whole process
could be better managed.
details:
Much of the algorithm is now documented at the top of the file
`heal_accounts.nim`
* Miscellaneous updates TBC
* Disentangled pivot2 module from snap
why:
Wrote as template on top of sync so it can be shared by fast and snap
sync.
* Renamed and relocated pivot sources
* Integrated `best_pivot` module into full and snap sync
why:
Full sync used an older version of `best_pivot`
* isolating download module from full sync
why;
might be shared with snap sync at a later stage
* Added inspect module
why:
Find dangling references for trie healing support.
details:
+ This patch set provides only the inspect module and some unit tests.
+ There are also extensive unit tests which need bulk data from the
`nimbus-eth1-blob` module.
* Alternative pivot finder
why:
Attempt to be faster on start up. Also tying to decouple pivot finder
somehow by providing different mechanisms (this one runs in `single`
mode.)
* Use inspect module for healing
details:
+ After some progress with account and storage data, the inspect facility
is used to find dangling links in the database to be filled nose-wise.
+ This is a crude attempt to cobble together functional elements. The
set up needs to be honed.
* fix scheduler to avoid starting dead peers
why:
Some peers drop out while in `sleepAsync()`. So extra `if` clauses
make sure that this event is detected early.
* Bug fixes causing crashes
details:
+ prettify.toPC():
int/intToStr() numeric range over/underflow
+ hexary_inspect.hexaryInspectPath():
take care of half initialised step with branch but missing index into
branch array
* improve handling of dropped peers in alternaive pivot finder
why:
Strange things may happen while querying data from the network.
Additional checks make sure that the state of other peers is updated
immediately.
* Update trace messages
* reorganise snap fetch & store schedule
* Re-implemented `hexaryFollow()` in a more general fashion
details:
+ New name for re-implemented `hexaryFollow()` is `hexaryPath()`
+ Renamed `rTreeFollow()` as `hexaryPath()`
why:
Returning similarly organised structures, the results of the
`hexaryPath()` functions become comparable when running over
the persistent and the in-memory databases.
* Added traversal functionality for persistent ChainDB
* Using `Account` values as re-packed Blob
* Repack samples as compressed data files
* Produce test data
details:
+ Can force pivot state root switch after minimal coverage.
+ For emulating certain network behaviour, downloading accounts stops for
a particular pivot state root if 30% (some static number) coverage is
reached. Following accounts are downloaded for a later pivot state root.
* Bump nim-stew
why:
Need fixed interval set
* Keep track of accumulated account ranges over all state roots
* Added comments and explanations to unit tests
* typo
* Extracted functionality into sub-modules for maintainability
* Setting SST bulk load as default in `accounts_db`
details:
+ currently, the same data are stored via rocksdb if available, and
the same via embedded `storage_type` with (non-standard) prefix 200
for time comparisons
+ fallback to normal `put()` unless rocksdb is accessible
* Provided common scheduler API, applied to `full` sync
* Use hexary trie as storage for proofs_db records
also:
+ Store metadata with account for keeping track of account state
+ add iterator over accounts
* Common scheduler API applied to `snap` sync
* Prepare for accounts bulk import
details:
+ Added some ad-hoc checks for proving accounts data received from the
snap/1 (will be replaced by proper database version when ready)
+ Added code that dumps some of the received snap/1 data into a file
(turned of by default, see `worker_desc.nim`)
* Error return in `persistBlocks()` on initial `VmState` roblem
why:
previously threw an exception
* Updated sync mode option
why:
using enum rather than bool => space for more
* Added sync mode `full`, re-factued legacy sync
also:
rebased
* Fix typo (crashes `pesistBlocks()` otherwise)
also:
rebase to master
* Reduce log ticker noise by suppressing duplicate messages
* Clarify staged queue overflow handling
why:
backtrack/re-org mode in `stageItem()` should be detected by both,
the global indicator or the work item where it might have moved into.
also:
rebased
* Relocated `IntervalSets` to nim-stew repo
* Accumulate accounts on temporary kv-DB
why:
Explore the data as returned from snap/1. Will be converted to a
`eth/db` next.
details:
Verify and accumulate per/state-root accounts downloaded via snap.
also:
Some unit tests
* Replace `Table` by `TrieDatabaseRef` for accounts accumulator
* update ticker statistics
details:
mean/variance based counter update
* allow persistent db for proved accounts
* rebase, and globally activate unit test
* fix statistics
* Using `IntervalSet` type data for `LeafRange`
* Updated log ticker
* Update to `eth67`
details:
Disabled by default, use `ENABLE_LEGACY_ETH66=0` to enable
No support for `Get/NodeData` dialogue via eth, anymore
* Dissolved fetch/common.nim
details;
the log/ticker part becomes ticker.nim
the interval range management is merged into fetch.nim
* Updated account scheduler
why:
The previous scheduler fetched each account once (for different state
roots.) The updated scheduler re-calibrates after a change of the state
root and potentially (until told otherwise) fetches all possible
accounts.
* Fix `high(P)` fringe cases in `IntervalSet` handling
why:
The `high(P)` value for a point type `P` cannot be represented with
half open intervals `[a,b)` for a,b points of `P`. So this single value
needs extra treatment which was slightly wrong.
* Updated docu/comments
also:
rebased
* Update scheduler
details:
Change the `pivot` management when creating new accounts lists. It is
strictly increasing (and wrapping around) depending on last updated
accounts list.
* Fix/recover download flag
why:
The fetch indicator used to control the data download somehow got
lost during re-org.
* Updated chronicles/logger topics
* Reorganised run state flags
why:
The original code used a pair of boolean flags `(stopped,stopThisState)`
which was translated to three states running, stoppedPending, and
stopped. It is currently not clear whether collapsing some states was
correct. So the original logic has been re-stored, albeit wrapped into
directives like `isStopped()` etc.
also:
Moving some function bodies in `worker.nim`
* Moved `reply_data.nim` and `validate_trienode.nim` to sub-directory `fetch_trie`
why:
Only used in `fetch_trie.nim`.
* Move `fetch_*` file and directory objects to `fetch` subdirectory
why:
Only used in `fetch.nim`
* Added start/stop and/or setup/release methods for all sub-modules
why:
good housekeeping
also:
updated getters/setters for ctrl states
updated trace messages
* Disentangle `collect` module from `reply_data`
why:
Now the module visible from `collect` for fetching data is `peer/fetch`
only.
* Merge `SnapPeerHunt` into `collect`
why:
This part needs to be known by `collect`, only
* rename collect => worker
* Dissolve `sync_fetch_xdesc` module into `common`
why:
Descriptor is only used in `common` and `fetch_trie`
* rename `snap/peer` directory => `snap/worker`
* rename `SnapSync` -> `Worker`, `SnapPeer` -> `WorkerBuddy`
* moved `snap/base_desc.nim` -> `snap/worker/worker_desc.nim`
* Unified opaque object ref naming in `worker_desc.nim`
details:
indicated my inheriting module (exactly one, always)
* Reorg SnapPeerBase descriptor, notably start/stop flags
details:
Instead of using three boolean flags startedFetch, stopped, and
stopThisState a single enum type is used with values SyncRunningOk,
SyncStopRequest, and SyncStopped.
* Restricting snap to eth66 and later
why:
Id-tracked request/response wire protocol can handle overlapped
responses when requests are sent in row.
* Align function names with source code file names
why:
Easier to reconcile when following the implemented logic.
* Update trace logging (want file locations)
why:
The macros previously used hid the relevant file location (when
`chroniclesLineNumbers` turned on.) It rather printed the file
location of the template that was wrapping `trace`.
* Use KeyedQueue table instead of sequence
why:
Quick access, easy configuration as LRU or FIFO with max entries
(currently LRU.)
* Dissolve `SnapPeerEx` object extension into `SnapPeer`
why;
It is logically cleaner and more obvious not to inherit from
`SnapPeerBase` but to specify opaque field object references of the
merged `SnapPeer` object. These can then be locally inherited.
* Dissolve `SnapSyncEx` object extension into `SnapSync`
why;
It is logically cleaner and more obvious not to inherit from
`SnapSyncEx` but to specify opaque field object references of
the `SnapPeer` object. These can then be locally inherited.
Also, in the re-factored code here the interface descriptor
`SnapSyncCtx` inherited `SnapSyncEx` which was sub-optimal (OO
inheritance makes it easier to work with call back functions.)
* new: time_helper, types
* new: path_desc
* new: base_desc
* Re-organised objects inheritance
why:
Previous code used macros to instantiate opaque object references. This
has been re-implemented with OO inheritance based logic.
* Normalised trace macros
* Using distinct types for Hash256 aliases
why:
Better control of the meaning of the hashes, all or the same format
caveat:
The protocol handler DSL used by eth66.nim and snap1.nim uses the
underlying type Hash256 and cannot handle the distinct alias in
rlp and chronicles/log macros. So Hash256 is used directly (does
not change readability as the type is clear by parameter names.)
* Use type name eth and snap (rather than snap1)
* Prettified snap/eth handler trace messages
* Regrouped sync sources
details:
Snap storage related sources are moved to common directory.
Option --new-sync renamed to --snap-sync
also:
Normalised logging for secondary/non-protocol handlers.
* Merge protocol wrapper files => protocol.nim
details:
Merge wrapper sync/protocol_ethxx.nim and sync/protocol_snapxx.nim
into single file snap/protocol.nim
* Comments cosmetics
* Similar start logic for blockchain_sync.nim and sync/snap.nim
* Renamed p2p/blockchain_sync.nim -> sync/fast.nim
why:
Accidentally wrapped into waitFor() directive with reviving jl/sync
branch.
also:
Decorate eth/66 and snap/1 protocol trace messages with protocol
type and version
* Squashed snap-sync-preview patch
why:
Providing end results makes it easier to have an overview.
Collected patch set comments are available as nimbus/sync/ChangeLog.md
in chronological order, oldest first.
* Removed some cruft and obsolete imports, normalised logging
* Activate wire protocol eth/66
and:
Disentangle protocol_eth66.nim from import sections
why:
Importing the protocol_eth66 module is not necessary. There is
no need to know too many details of the underlying wire protocol. All
that is needed will be exported by blockchain_sync.nim.
* fixes, and rebase
* Update nimbus/p2p/blockchain_sync.nim
Co-authored-by: Kim De Mey <kim.demey@gmail.com>
* Fixes and rebase
Co-authored-by: Kim De Mey <kim.demey@gmail.com>
The format is reasonably useful and not too large, when looking at the
behaviour of sync processes. It doesn't try to show all the details of
packets, just something at a useful level of detail to see what's going on.
The consistent presentation has proven helpful too, e.g. when grepping.
Signed-off-by: Jamie Lokier <jamie@shareable.org>
This constant shouldn't be used outside `protocol_eth65`.
When we support multiple `eth/NN` versions side by side, or even just have
multiple code files, there's a risk some code would import just one of the
files (e.g. `protocol_eth65`), use `protocolVersion`, and incorrectly act as
though that version is the one active on the node.
In fact that happened, and now it can't happen. Other code needs to query the
`EtheruemNode` to find what versions are really active.
Signed-off-by: Jamie Lokier <jamie@shareable.org>
Turns out `{.rlpInline.}` doesn't do anything.
It's documented but not implemented.
Due to this, whenever a peer sent us a `NewBlock` message, we had an RLP
decoding error processing it, and disconnected the peer thinking it was the
peer's error.
These messages are sent often by good peers, so whenever we connected to a
really good peer, we'd end up disconnecting from it due to this.
Because a block body is a list of transactions, the parse errors looked
suspiciously like EIP-2718/2976/2930/1559 typed transaction RLP errors.
But it was a failure to parse `BlockBody` inline.
Conveniently, the `EthBlock` type defined for another reason is encoded exactly
the way `NewBlockAnnounce` needs to be, so we can reuse that type.
This didn't stand out before updating to `eth/65`, because with old protocols
we tend to only connect to old peers, which may be out of date themselves and
have no new blocks to send. Also, we didn't really investigate occasional
disconnects before, we assumed they're just part of P2P life.
Signed-off-by: Jamie Lokier <jamie@shareable.org>
This patch adds the `eth/65` protocol, documented at
https://github.com/ethereum/devp2p/blob/master/caps/eth.md.
This is an intentionally simple patch, designed to not break, change or depend
on other functionality much, so that the "_old_ sync" methods can be run
usefully again and observed. This patch isn't "new sync" (a different set of
sync algorithms), but it is one of the foundations.
For a while now Nimbus Eth1 only supported protocol `eth/63`. But that's
obsolete, and very few nodes still support it. This meant Nimbus Eth1 would
make slow progress trying to sync, as most up to date clients rejected it.
The current specification is `eth/66`, and the functionality we really need is
in `eth/64`.
So why `eth/65`?
- `eth/64` is essential because of the `forkId` feature. But `eth/64` is on
its way out as well. Many clients, especially the most up to date Geth
running the current hard-forks (Berlin/London) don't talk `eth/64` any more.
- `eth/66` is the current specification, but some clients don't talk `eth/66`
yet. We'd like to have the best peer connectivity during tests, and
currently everything that talks `eth/66` also talks `eth/65`.
- Nimbus Eth1 RLPx only talks one version at a time. (Without changes to the
RLPx module. When those go in we'll add `eth/64..eth/66` for greater peer
reach and testing the `eth/66` behaviour. For simplicity and decoupling,
this patch contains just one version, the most useful.)
What are `eth/64` and `eth/65`?
- `eth/64` (EIP-2364) added `forkId` which allows nodes to distinguish between
Ethereum (ETH) and Ethereum Classic (ETC) blockchains, which share the same
genesis block. `forkId` also protects the system when a new hard fork is
going to be rolled out, by blocking interaction with out of date nodes. The
feature is required nowadays.
We send the right details to allow connection (this has been tested a lot),
but don't apply the full validation rules of EIP-2124/EIP-2364 in this patch.
It's to keep this patch simple (in its effects) and because those rules have
consequences best tested separately. In practice the other node will reject
us when we would reject it, so this is ok for testing, as long as it doesn't
get seriously deployed.
- `eth/65` added more efficient transaction pool methods.
- Then a later version of `eth/65` (without a new number) added typed
transactions, described in [EIP-2976](https://eips.ethereum.org/EIPS/eip-2976).
Why it's moved to `nimbus-eth1`:
- Mainly because `eth/64` onwards depend on the current state of block
synchronisation, as well as the blockchain's sequence of hard-fork block
numbers, both of which are part of `nimbus-eth1` run-time state. These
aren't available to pure `nim-eth` code. Although it would be possible to
add an API to let `nimbus-eth1` set these numbers, there isn't any point
because the protocol would still only be useful to `nimbus-eth1`.
Signed-off-by: Jamie Lokier <jamie@shareable.org>
jangko
Jordan Hrycaj
Jacek Sieka
bmoo
Jamie Lokier