Skip to content

core, core/state: instrument BAL slow-block metrics#34861

Merged
jwasinger merged 6 commits into
ethereum:bal-devnet-3from
CPerezz:bal-slow-block-metrics
May 12, 2026
Merged

core, core/state: instrument BAL slow-block metrics#34861
jwasinger merged 6 commits into
ethereum:bal-devnet-3from
CPerezz:bal-slow-block-metrics

Conversation

@CPerezz
Copy link
Copy Markdown
Contributor

@CPerezz CPerezz commented May 1, 2026 •
edited
Loading

A first attempt to get the entire slow_blocks to be BAL&parallel-exec-aware.

  • Sequential-path output is unchanged; the bal JSON field is omitempty.
  • All new types are pure-data snapshots (value receivers, no shared mutation), so they're safe to copy across goroutine boundaries in the parallel pipeline.

@CPerezz CPerezz requested a review from rjl493456442 as a code owner May 1, 2026 13:33
@CPerezz
Copy link
Copy Markdown
Contributor Author

CPerezz commented May 3, 2026

cc: @jwasinger this should do the job.

LMK what's your take :)

jwasinger
jwasinger reviewed May 3, 2026
View reviewed changes
Comment thread core/state/bal_state_transition.go Outdated
Comment on lines +56 to +57
accountReadNS atomic.Int64
storageReadNS atomic.Int64
Copy link
Copy Markdown
Contributor

@jwasinger jwasinger May 3, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

db read operations should be highly concurrent here. is the aggregate time from all operations a useful metric to track?

Copy link
Copy Markdown
Contributor Author

@CPerezz CPerezz May 4, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

hmmmm. fair point it may be not. We had it in the original spec IIRC. And I think it was just kept? It can show somewhat valuable things like any tradeoffs on account vs. storage read optimizations we might be missing (?).

Unsure. What's your take? I also wonder what @rjl493456442 thinks here.

jwasinger
jwasinger reviewed May 3, 2026
View reviewed changes
Comment thread core/blockchain.go Outdated
@jwasinger
Copy link
Copy Markdown
Contributor

jwasinger commented May 3, 2026 •
edited
Loading

Thanks for this! Mostly looks good. I will push some small cleanups to it tomorrow and then merge it.

@jwasinger
Copy link
Copy Markdown
Contributor

jwasinger commented May 5, 2026

Thanks for this! Mostly looks good. I will push some small cleanups to it tomorrow and then merge it.

I spoke too soon.. ended up getting sick from the plane flight and/or airports on the trip back home.

jwasinger
jwasinger reviewed May 5, 2026
View reviewed changes
Comment thread core/state/bal_state_transition.go Outdated
jwasinger
jwasinger reviewed May 5, 2026
View reviewed changes
Comment thread core/state/statedb.go Outdated
jwasinger
jwasinger reviewed May 5, 2026
View reviewed changes
Comment thread core/state/statedb.go Outdated
@jwasinger jwasinger mentioned this pull request May 6, 2026
Merged
jwasinger added a commit that referenced this pull request May 6, 2026
@jwasinger @CPerezz
cmd,core,tests: introduce new BAL execution flags, log BAL slow block…
697fe91 
…s, surface more metrics including prefetcher time (#34892)

Adapts some of the changes from
#34861 . Some other metrics
which are recorded manually during execution in that PR, but can be
deduced from the BAL are TBD.

I've added two bal feature flags:

* `--bal.prefetchworkers <uint>`: this tunes the number of concurrent
go-routines that will be used to perform state fetching tasks by the BAL
prefetcher. Default is `runtime.NumCPUs`, the current behavior in
`bal-devnet-3`.
* `--bal.blockingprefetch`: If set, state prefetching will block the
execution of transactions and state root update.

---------

Co-authored-by: CPerezz <cperezz19@pm.me>
@CPerezz CPerezz force-pushed the bal-slow-block-metrics branch from d0005c1 to 3f5e27e Compare May 6, 2026 22:14
@CPerezz
Copy link
Copy Markdown
Contributor Author

CPerezz commented May 6, 2026

@jwasinger this should be ready!

jwasinger
jwasinger reviewed May 7, 2026
View reviewed changes
Comment thread core/state/reader.go Outdated
jwasinger
jwasinger reviewed May 7, 2026
View reviewed changes
Comment thread core/state/state_object.go
jwasinger
jwasinger reviewed May 7, 2026
View reviewed changes
Comment thread core/state/statedb.go Outdated
jwasinger
jwasinger reviewed May 7, 2026
View reviewed changes
Comment thread core/state/statedb.go
jwasinger
jwasinger reviewed May 7, 2026
View reviewed changes
Comment thread core/blockchain.go Outdated
jwasinger
jwasinger reviewed May 11, 2026
View reviewed changes
Comment thread core/state/bal_state_transition_test.go Outdated
jwasinger
jwasinger reviewed May 11, 2026
View reviewed changes
Comment thread core/state/reader.go Outdated
jwasinger
jwasinger reviewed May 11, 2026
View reviewed changes
Comment thread core/state/reader.go
CPerezz and others added 6 commits May 12, 2026 10:06
@CPerezz @jwasinger
core, core/state: instrument BAL slow-block metrics
fd76921 
Populates per-block state read/write counts in slow-block JSON for BAL
blocks (which ethereum#34892 left as TBD), and adds reader-level read timing.
Builds on top of bal-devnet-3 — most of the PR's earlier slow-block log
infrastructure was adapted into upstream by that commit, so this change
is now scoped to the metric population that the BAL alone can derive.

- BAL helpers: BlockAccessList.{UniqueAccountCount, UniqueStorageSlotCount,
  WrittenCounts}. WrittenCounts walks the BAL once and returns the
  block-aggregate write counts.

- Reader-level read timing: *reader times all synchronous Account/Storage/
  Code/CodeSize calls via atomic counters; exposed via ReadTimes()
  ReadDurations and the new state.ReadTimer interface. Replaces StateDB-
  level AccountReads/StorageReads/CodeReads tracking (the StateDB shouldn't
  time its dependencies — the reader is where the I/O happens).

- Reader-level code-load dedup: *reader.codeLoaded sync.Map records the
  first-seen byte length per address; CodeLoads() returns (count, bytes).
  Exposed via state.CodeLoadTracker. Replaces StateDB CodeLoaded/
  CodeLoadBytes tracking and the SnapshotCodeLoads aggregation pattern.

- BALStateTransition: caches BlockAccessList.WrittenCounts() once at
  construction; tracks accountDeleted/storageDeleted atomics for the
  parallel root-pass (the BAL alone can't distinguish a selfdestruct from
  a balance/nonce reset). Exposes Deletions() DeletionCounts. Drops the
  older accountUpdated/storageUpdated/codeUpdated/codeUpdateBytes counters
  (now derived from WrittenCounts).

- BAL block stats path (blockchain.go): populates StateCounts directly —
  AccountUpdated = WrittenCounts.Accounts - Deletions.Accounts (same for
  storage). AccountLoaded/StorageLoaded come from BAL. CodeLoaded/
  CodeLoadBytes come from the shared *reader (deduplicated across phase
  StateDBs naturally because they share one reader instance).

- Non-BAL block stats path: read durations come from the reader; counts
  from StateDB fields. StorageUpdated/StorageDeleted unified to int width.

- Hard type assertions: state.ReadTimer / state.CodeLoadTracker /
  state.ReaderStater consumers use direct casts (no silent zero
  fallback) — every Reader chain in production satisfies these
  interfaces.

- Meter alignment: account/storage Updated meters subtract Deletions to
  avoid double-reporting blocks under both Update and Delete dashboards.
@CPerezz @jwasinger
core, core/state: address slow-block review feedback
4183a70 
- Pre-BAL path keeps StateDB synchronous read-time accumulators
  (AccountReads, StorageReads, CodeReads, CodeLoaded, CodeLoadBytes)
  so Execution = ptime - reads stays well-formed under single-thread
  execution.
- BAL path drops aggregate reader read-times; under parallel workers
  they sum across goroutines and aren't a wall-clock proxy.
- Delete dead PrefetchReadTimes/WaitPrefetch forwarders on *reader and
  the now-unused ReadTimer/ReadDurations scaffolding.
- Add regression test for EIP-7702 delegation clear: empty []byte code
  in the BAL must reset CodeHash to EmptyCodeHash.
@CPerezz @jwasinger
core/state: restore BALStateTransition descriptor comments
e4f2727 
Stripped by 3f5e27e alongside the StateDB ones. The struct doc and
per-field descriptors (rootHash, diffs, prestates, tries) come back.
@CPerezz @jwasinger
core, core/state: fold prefetcher elapsed into StateReadMs for BAL bl…
659af06 
…ocks

For BAL blocks AccountReads/StorageReads/CodeReads stay zero by design,
so the slow-block JSON state_read_ms read 0 — misleading to cross-client
consumers. Sum s.Prefetch (the prefetcher's elapsed time) into the field;
it's already zero on the pre-BAL path so the formula is unchanged there.

Also gofmt -w on reader.go and reader_eip_7928.go (trailing blank lines
left over from earlier cleanup).
@CPerezz @jwasinger
core/state: drop unused reader forwarders and EIP-7702 BAL test
f615a62 
The Account and Storage methods on *reader were pure forwarders to the
embedded StateReader; Go's method promotion makes them redundant.

Remove the EIP-7702 delegation-clear regression test; coverage is
deferred to the execution-spec test suite, which already exercises the
BAL apply path via tests/block_test_util.go.
@jwasinger
add back some func/struct descriptor comments
3eafb4e
@jwasinger jwasinger force-pushed the bal-slow-block-metrics branch from 9bd26a3 to 3eafb4e Compare May 12, 2026 14:07
@jwasinger jwasinger merged commit ac36e48 into ethereum:bal-devnet-3 May 12, 2026
2 checks passed
@CPerezz CPerezz deleted the bal-slow-block-metrics branch May 12, 2026 15:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jwasinger jwasinger jwasinger left review comments

@rjl493456442 rjl493456442 Awaiting requested review from rjl493456442 rjl493456442 is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@CPerezz @jwasinger