Rongsong Shen
d26ff8870d
The buglog was one 27K-line JSON file whose fields covered only symptom,
root cause and fix, as free text: 2% of entries named the code version a
bug was found on, 12% the build profile, 36% a way to reproduce, and 58% of
the file was OpenWolf hook auto-detections ("Missing guard clause",
"Function not marked async" on .md files).
- .claude/skills/buglog/bug-record.md: the record standard. Eight
sections (Summary as the title, Description, Environment, Reproduction;
Root cause, Fix, Verification, References once fixed), each with the
content it must contain -- e.g. Environment needs the OS, the libc the
failing binary loads (from its ELF interpreter: glibc 2.42 in the nix
shell vs 2.40 on the host), toolchain, build profile and commit.
- .wolf/bugs/<id>.md: 500 records migrated from buglog.json; every field
verified to round-trip. They carry legacy: true and an explicit "not
recorded" where the old log had no such field; component derived from
file paths, severity from keywords (else unknown), fix_commits from
hashes in status_note/fix that git resolves (91 of 463 fixed).
- .wolf/buglog-archive.json: the 710 hook auto-detections, verbatim.
- .wolf/buglog.json is now an index generated from the records, keeping
the fields OpenWolf's hooks and `openwolf bug search` read.
- scripts/buglog.py rewritten: new (template + Environment collected from
the machine and --binary), check, status (refuses a status the record
doesn't support; fixed needs Root cause, Fix, Verification and
--commit hashes git resolves), note, recur, set, show, list, triage,
env, index. 29 unittest cases in scripts/tests/test_buglog.py.
- scripts/git-hooks/pre-commit runs `check --staged`: incomplete staged
records, a legacy record hand-changed to fixed without the fix-time
sections, a new record claiming legacy: true, or a stale index refuse
the commit.
- scripts/gitea-sync.py reads the records: issue body = the record's
sections plus linked fix commits (origin remote, GITEA_COMMIT_REPO to
override), and a body hash so an issue's text is refreshed when the
record changes.
- bug-1866 rewritten to the full standard as the worked example.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
38 lines
3.6 KiB
Markdown
38 lines
3.6 KiB
Markdown
---
|
|
id: "bug-991"
|
|
status: "fixed"
|
|
status_updated: "2026-09-24"
|
|
status_note: ""
|
|
component: "controller"
|
|
severity: "unknown"
|
|
timestamp: "2026-06-22T16:30:00.000Z"
|
|
last_seen: "2026-09-19T09:00:00Z"
|
|
occurrences: 1
|
|
files: ["src/controller/ha.c (ha_apply_term_entry_via_library ~2115; batch ~2205; single-entry ~2225)"]
|
|
tags: ["ha", "controller", "wal-catchup", "applied_lsn", "commit-stall", "off-by-one", "term-safety", "bug-991", "streamer-skip"]
|
|
related: ["bug-851", "bug-850", "bug-848", "bug-924", "bug-960"]
|
|
legacy: true
|
|
---
|
|
|
|
# bug-991: Standby WAL catch-up advanced applied_lsn to lsn+1 (next-to-apply) instead of lsn (last-held)
|
|
|
|
## Description
|
|
Standby WAL catch-up advanced applied_lsn to lsn+1 (next-to-apply) instead of lsn (last-held). The standby then advertised one LSN too high; the leader's catch-up streamer streamed (lsn+1, committed] and SKIPPED the real next-needed entry lsn+1, wedging the library logOk gate forever (no WalAck -> match_index never reaches -> raft commit_index frozen -> committed_lsn pinned at 1). Catch-up-path-specific; live repl_broadcast (absolute-LSN, no skip) commits fine, so failure was nondeterministic.
|
|
|
|
## Environment
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Reproduction
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Root cause
|
|
applied_lsn carried TWO conflicting semantics. Recovery (=durable_floor), checkpoint (=checkpoint_lsn), init (=0) and the leader's catch-up streamer + ha_evaluate_repl_action all treat it as 'last contiguously-held LSN' and stream (standby_lsn, committed]. But the three live/catch-up apply setters advanced to entry->lsn + 1 ('next-to-apply'). After a standby applied catch-up entry N it advertised N+1, so the leader's next stream began at N+2 and skipped N+1 -> logOk reject -> no WalAck.
|
|
|
|
## Fix
|
|
fix-1: the three live-apply setters now advance applied_lsn to the LAST-HELD LSN (e->lsn / last_good_lsn / entry->lsn), not +1, so applied_lsn means 'last contiguously-held LSN' everywhere. Reader audit (all consistent with last-held, no +1 reader needed): state report (advertised LSN, ha.c:2461) now matches streamer contract; convergence reported_lsn>=committed_lsn (2536) correct; snapshot/stream boundary standby_lsn+1<oldest_wal_lsn (2501) already assumed last-held; vote/prevote (5106/5183) feed applied_lsn only to the TAG fold, LSN axis overridden by ha_candidate_advertise_lsn(durable_floor); recovery ha_sync_applied_lsn_from_recovery already last-held. fix-B (source advertised LSN from raft.log_len) NOT landed: on the production catch-up road every streamed entry is term-carrying -> routed through the library, so raft.log_len already == applied_lsn after fix-1; raft.log_len legitimately differs from applied_lsn on the legacy term-0 direct-apply and recovery paths, so re-sourcing would regress bug-848 incremental catch-up -> fix-1 alone delivers the value the streamer+logOk need. Unit RED->GREEN: test_catchup_applied_lsn_is_last_held_no_skip (Expected 1 Was 2 pre-fix, PASS post-fix). 7 legacy test_wal_catchup_* expected values updated lsn+1 -> last-held. NOTE: the LIVE clean_failover_preservation STRICT_TERM commit=5 target was NOT reachable in the sandbox: committed_lsn stalls at 1 IDENTICALLY with this fix stashed (baseline experiment) -> that stall is a SEPARATE upstream blocker (library raft.log_len / commit-advance not advancing past the own-term noop; ha-term-safety task 3.1 area), NOT the bug-991 applied_lsn skip. waldump confirms all 3 members hold LSN 1..5 byte-identical (term 1) -> the replication half (which this fix corrects) is sound.
|
|
|
|
## Verification
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## References
|
|
- Related: bug-851, bug-850, bug-848, bug-924, bug-960
|