Skip to content

fix(cache,da): drop finalized snapshot entries on restore and drain DA subscription channel - #3385

Merged
tac0turtle merged 3 commits into
mainfrom
marko/cache-restore-filter-sub-drain
Jul 16, 2026
Merged

fix(cache,da): drop finalized snapshot entries on restore and drain DA subscription channel#3385
tac0turtle merged 3 commits into
mainfrom
marko/cache-restore-filter-sub-drain

Conversation

@tac0turtle

@tac0turtletac0turtle commented Jul 15, 2026

Copy link
Copy Markdown
Contributor

Overview

Two memory fixes for long-running nodes, found while investigating a user report of sustained memory growth (+201 MiB/min) and OOM kills on syncing nodes.

1. Cache restore leaks stale placeholders across crash/restart cycles

The DA-inclusion snapshot is only persisted on graceful shutdown (Components.Stop). After a crash (e.g. OOM kill, SIGKILL), the node restores a snapshot from the last graceful shutdown, which can contain heights already at or below the persisted DAIncludedHeight watermark. Since:

  • the inclusion loop only calls DeleteHeight for heights it advances through, and
  • IsHeightDAIncluded short-circuits true below the watermark so those entries are never consulted,

restored placeholders below the watermark were never evicted, and were re-persisted on the next graceful save — so the snapshot grew monotonically across crash/restart cycles.

Cache.RestoreFromStore now takes the finalized height and skips entries at or below it. Skipped entries still contribute to maxDAHeight, so DaHeight() semantics are unchanged. manager.RestoreFromStore reads DAIncludedHeightKey and passes it through; a read error falls back to unfiltered restore instead of failing startup.

The pre-existing comment in TestManager_SaveAndRestoreFromStore already described this as the intended behavior ("The cache entry is not restored — this is correct and intentional") — the code just didn't do it. It now does, and the test asserts it.

2. DA subscription goroutine leak on reconnect

The client.Subscribe wrapper goroutine exited on ctx cancellation without draining the underlying jsonrpc channel. The go-jsonrpc delivery goroutine blocked on send into that channel never observed the cancellation and never closed its channel — leaking one goroutine per watchdog reconnect (observed in the field as Subscribe.func1 goroutines doubling). The wrapper now drains the raw channel on exit, matching what its own doc comment previously required of callers.

Testing

  • New TestCache_RestoreFromStore_SkipsFinalizedEntries covering partial and fully-stale snapshots
  • TestManager_SaveAndRestoreFromStore now asserts finalized heights are not restored
  • go test ./block/..., go vet, golangci-lint all clean

Note: this does not address the primary OOM driver from the same investigation (unbounded pendingEvents buffering when DA catchup outpaces execution) — that needs a separate backpressure change.

🤖 Generated with Claude Code

Summary by CodeRabbit

  • Bug Fixes

    • Prevented finalized cache entries from being restored as active placeholders after restart.
    • Preserved the correct maximum data-availability height while removing obsolete entries.
    • Improved cache restoration when persisted metadata is unavailable.
    • Prevented subscription shutdown and reconnection scenarios from leaving message delivery blocked.
  • Reliability

    • Improved consistency of restored header and data availability state across restarts.

…A subscription channel
Two memory fixes for long-running nodes:
Cache restore: the DA-inclusion snapshot is only written on graceful
shutdown, so after a crash (e.g. OOM kill) the restored snapshot can
contain heights already below the persisted DA-included watermark. The
inclusion loop never evicts below its watermark, so those placeholder
entries leaked for the process lifetime and were re-persisted on every
subsequent save, growing the snapshot monotonically across crash/restart
cycles. RestoreFromStore now skips entries at or below the persisted
DAIncludedHeight; skipped entries still seed maxDAHeight so DaHeight()
is unchanged.
DA subscription: the Subscribe wrapper goroutine exited on ctx
cancellation without draining the underlying jsonrpc channel, leaving
the go-jsonrpc delivery goroutine blocked on send — it never observed
the cancellation and never closed its channel, leaking one goroutine
per watchdog reconnect. The wrapper now drains the raw channel on exit.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@coderabbitai

coderabbitaiBot commented Jul 15, 2026

Copy link
Copy Markdown
Contributor

Review Change Stack

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: ee127894-6385-4068-92d9-4d2930c23502

📥 Commits

Reviewing files that changed from the base of the PR and between 19d3a86 and 896871f.

📒 Files selected for processing (5)
  • block/internal/cache/generic_cache.go
  • block/internal/cache/generic_cache_test.go
  • block/internal/cache/manager.go
  • block/internal/cache/manager_test.go
  • block/internal/da/client.go

📝 Walkthrough

Walkthrough

Cache restoration now accepts a finalized-height watermark, skips finalized placeholders while retaining DA height, and passes persisted metadata through manager restoration. DA subscription shutdown now drains the underlying JSON-RPC channel.

Changes

Watermarked cache restoration

Layer / File(s)Summary
Finalized-aware snapshot restoration
block/internal/cache/generic_cache.go, block/internal/cache/generic_cache_test.go
RestoreFromStore preserves the maximum decoded DA height while skipping snapshot entries at or below the finalized height; tests cover stale and partially finalized snapshots.
Manager integration and regression coverage
block/internal/cache/manager.go, block/internal/cache/manager_test.go, block/internal/cache/generic_cache_test.go
Manager restoration reads the persisted DA-included height and passes it to both caches; restore tests use the updated signature and verify finalized entries are absent.

DA subscription shutdown

Layer / File(s)Summary
Subscription channel draining
block/internal/da/client.go
Subscribe drains the raw JSON-RPC channel in a deferred shutdown loop and updates its lifecycle documentation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CacheManager
participant Store
participant Cache
CacheManager->>Store: read DAIncludedHeightKey
CacheManager->>Cache: restore with finalized height
Cache->>Cache: update DA height from snapshot
Cache->>Cache: install only non-finalized placeholders
Loading
sequenceDiagram
participant Caller
participant Subscribe
participant JSONRPC
Caller->>Subscribe: cancel context
Subscribe->>JSONRPC: drain raw channel
JSONRPC-->>Subscribe: close channel
Loading

Possibly related PRs

Suggested reviewers:julienrbrt

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check nameStatusExplanation
Title check✅ PassedThe title accurately summarizes both main fixes: skipping finalized cache entries on restore and draining the DA subscription channel.
Description check✅ PassedThe description matches the template with a clear Overview, rationale, and testing section, and it explains both fixes in enough detail.
Docstring Coverage✅ PassedNo functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check✅ PassedCheck skipped because no linked issues were found for this pull request.
Out of Scope Changes check✅ PassedCheck skipped because no linked issues were found for this pull request.
✨ Finishing Touches
📝 Generate docstrings
  • Create stacked PR
  • Commit on current branch
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch marko/cache-restore-filter-sub-drain

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@github-actions

github-actionsBot commented Jul 15, 2026

Copy link
Copy Markdown
Contributor

The latest Buf updates on your PR. Results from workflow CI / buf-check (pull_request).

BuildFormatLintBreakingUpdated (UTC)
✅ passed⏩ skipped✅ passed⏩ skippedJul 16, 2026, 10:29 AM

@codecov

codecovBot commented Jul 15, 2026

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 50.00000% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 62.32%. Comparing base (19d3a86) to head (896871f).

Files with missing linesPatch %Lines
block/internal/cache/manager.go16.66%2 Missing and 3 partials ⚠️
block/internal/da/client.go50.00%0 Missing and 1 partial ⚠️
Additional details and impacted files
@@ Coverage Diff @@## main #3385 +/- ##
==========================================
- Coverage 62.33% 62.32% -0.01% 
==========================================
Files 120 120 Lines 13444 13452 +8 ==========================================
+ Hits 8380 8384 +4 - Misses 4124 4126 +2 - Partials 940 942 +2 
FlagCoverage Δ
combined62.32% <50.00%> (-0.01%)⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@tac0turtle
tac0turtle merged commit b4b895b into mainJul 16, 2026
28 checks passed
@tac0turtle
tac0turtle deleted the marko/cache-restore-filter-sub-drain branch July 16, 2026 10:52
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

@tac0turtle