4 Commits

Include renames

Author	SHA1	Message	Date
Chris Lu	7063b3e14c	s3 lifecycle: bound the daily-replay pass so a quiet cluster stops wedging the job (#10578) ... * s3 lifecycle: bound the daily-replay subscription at the pass boundary A pass opens one meta-log subscription and 16 shard drains, then waits on all of them. Nothing told the subscription where the pass ends, so the only exit was the fan-out spotting an event past runNow — i.e. some unrelated write landing under /buckets after the pass started. On a cluster that goes quiet the reader parks in Recv, every shard drain starves on an empty channel, and Run never returns. The job sits at stage "starting" with the executor slot held and no log line, so expiry stops cluster-wide until someone restarts the worker. The pass covers (globalStartTsNs, runNow], so say that: UntilNs on the subscribe request makes the filer end the stream once it has shipped that range. The reader then closes the event channel on the way out, which is what unblocks the fan-out and the drains when the stream finishes on its own rather than by cancellation. Same fix retires the other silent hang: a reader that failed early (subscribe error, stream error) also left every drain waiting forever. * s3 lifecycle: keep a halted shard from starving the shared fan-out A drain that halts mid-stream (BLOCKED / RETRY_LATER / an RPC error on dispatch) returns while the fan-out is still routing that shard's events. After 256 of them the per-shard buffer is full and the fan-out blocks on the send, so no other shard sees another event. Run's WaitGroup never drains, and the teardown that would cancel the reader sits behind that wait — the pass wedges exactly like an idle subscription did, with one S3 hiccup as the trigger. Keep discarding the channel after runShard returns. The events are past this shard's saved cursor and get re-scanned next pass anyway. * s3 lifecycle: assert the starved shard actually made progress The fan-out test only checked that Run returned, which a version that quietly dropped the second shard's events would also satisfy. Assert the dispatch landed and the cursor moved. recordingClient gains a per-object outcome map: the two shards dispatch from separate goroutines, so pinning BLOCKED by call index was a race waiting to pick the wrong shard. * s3 lifecycle: fail the pass when the shared subscription dies Closing the event channel on reader exit is what unblocks the shard drains, but it also means a subscribe that never opened, or a stream that broke mid-pass, now ends every drain cleanly. Run logged that at V(2) and returned the shard result — so a filer failure produced a green lifecycle job that had processed nothing. Surface it as the pass error. Cursors still hold what was processed and tomorrow resumes there; what changes is that the job stops claiming success. Cancellation has to stay a non-error — the shell driver's -runtime cap is a truncated pass, not a failed one — and a canceled gRPC stream arrives as a status code, not a wrapped context.Canceled, so isCanceled checks both forms the way the rest of the tree does. * s3 lifecycle: decide reader cancellation by intent, not status code A stream we cancel and a stream the filer cancels both arrive as codes.Canceled, so classifying the reader's exit by its error let a truncated pass report success whenever the failure happened to carry a cancellation status. Intent is knowable exactly, so read that instead: the pass stops on purpose only when the caller's context ended (the shell driver's -runtime cap) or the fan-out hit the pass boundary itself. Everything else is a broken subscription and fails the pass. TestRun_ServerSideCancelFailsThePass and TestRun_CappedPassIsNotAFailure are the same codes.Canceled from the reader with opposite verdicts — the pair only passes because the decision no longer looks at the error. * s3 lifecycle: time out a subscription that stops delivering UntilNs ends a healthy stream and gRPC keepalive catches a dead connection, but neither reaches a filer that keeps answering pings while its handler has stopped producing. The pass would wait on that forever, since s3_lifecycle is the one job type with no execution timeout. Bound the wait for each response at 20 minutes, and opt into the filer's idle heartbeats so a caught-up stream proves liveness instead of looking stalled. The default sits above the filer's 15-minute metadata-gap recovery budget, so a subscriber legitimately parked on a gap is never mistaken for a stalled one. Recv is only interruptible by killing the RPC, so it moves to its own goroutine behind a per-response deadline. The timer covers only the wait on the filer — dispatch to Events happens outside it, so a slow consumer can't trip the watchdog. Approach and the 20-minute figure are from #10577 by way of comparing the two fixes; the wiring differs because the reader here ends the pass by closing its event channel rather than cancelling the fan-out. * s3 lifecycle: trim the comments added by this branch Keep the non-obvious why, drop the prose restating what the code says. * s3 lifecycle: snapshot reader intent where the reader stops Sampling ctx.Err() during teardown reads it after the drains and cursor saves have run. A reader that failed while the deadline was still live, on a pass whose teardown then outlives that deadline, was classified as an intentional stop and reported success. Sampling earlier in Run is not the fix either: before the shard wait, a legitimately capped pass has not reached its deadline yet and would be misclassified the other way. Intent belongs where the reader actually stops, so the reader goroutine records it next to the error it returns. Reported by greptile on #10578. * s3 lifecycle: cover the worker-dispatched pass with nothing due The e2e suite drives the shell command in 14 of 15 files; the one test on the real admin->worker path backdates an object, so its own delete pushes a meta-log event past the pass boundary and ends the pass. The branch where a pass has nothing to dispatch was never exercised through the worker. Cover it, asserting the pass returns on its own: no admin cancellation, and the executor slot free for the next one. This is not a regression test for the wedge. A pass used to end when any write landed past its boundary, and on a shared test cluster something usually does — the whole suite passes on the unfixed build, verified. The deterministic guards stay the dailyrun unit tests; this one would catch a pass that hangs unconditionally.	2026-08-05 08:41:37 -07:00
Chris Lu	47b491b53c	mount: version open file handles by filer log position (#10403) ... * filer: stamp a log position on lookup and remote-cache responses Metadata events are logged after their store write and stamped with the filer clock. Reading that clock before serving an entry therefore gives a timestamp with a causal guarantee: every event at or below it is reflected in the returned entry. Clients caching filer state can use it as the entry's version to order the response against subscription events, including events committed before the call but delivered after it. * mount: version open file handles by filer log position A subscription event refreshing an open handle did a second lookup; a transient failure left the handle pinned to its old entry with no retry, since the subscription cursor had already advanced. The deeper problem is ordering: the handle is a cache written by three unordered channels — the async invalidation worker, local mutation acks, and open-time lookups — and overwriting cached state safely requires knowing which write is newer. The filer log timestamp is that order, and it now travels with every value instead of being derived out of band. Events carry it natively; lookup and remote-cache responses carry the log position stamped before the serving read; mutation acks carry it in their returned event; and the local store pairs each read with a version cursor advanced under the same lock as the store write. Each handle records the version its entry reflects, and one rule replaces the per-site reasoning: state at or below the handle's version is old news and must not be installed. The invalidation itself applies the event's own entry — no lookup, so no transient-failure window — except under a cached parent, where the store entry is the ordered merge of the event and anything applied since, and its version outranks the event's. An uncached parent receives no store writes, so a hit there would be a stale leftover masking the event. A vacated path (delete, rename away) keeps the last entry so unlinked-but-open reads still work. Directory builds version the completed directory at the listing snapshot and re-invalidate buffered events at that version, since their mid-build refresh ran against an incomplete store. The tests replay every race this replaces machinery for: rollback of a newer local flush (queued, cached, and read-through), stale leftovers under uncached parents, the build window including abort, handles opened after an event was queued, events landing mid-lookup, and undelivered events at remote-cache time across a filer failover. * filer: serialize the log position fence with mutations, stamp mutation acks The fence stamped before an unlocked entry read could precede state the read returned: a mutation writes storage first and assigns its event timestamp only at notify time, so a lookup racing that window handed the mount an entry newer than its fence, and the event's later delivery looked like fresh news — destroying dirty pages for a change the handle already had. The mutation handlers already hold an exclusive per-path lock across read, write, and notify; the lookup and remote-cache reads now take it shared around the stamp and the read, making the fence exact: everything at or below it is in the entry, nothing above it is. A no-change update returns success without an event, leaving the mount nothing to fence with even though the response confirms current state. Create and update acks now carry a log position stamped under the same lock, and the mount falls back to it whenever the ack has no event. Also regenerate the VT marshalers, which the earlier generation missed: without them a VT round-trip silently zeroed every log position. * java: sync filer.proto * mount: scope store versions to what they vouch for; atomic handle install The store's version cursor claimed too much. Advanced by local mutation acks and directory listing snapshots, it inflated the version of store reads for unrelated paths whose events the subscription still owed, and those events were then fenced out permanently. The cursor now tracks subscription progress only — events arrive in log order, so everything at or below it has been delivered for every path — and a completed listing records its snapshot as a per-directory floor instead of a global claim. Local acks never touch it: they version their own handle directly. Buffered build events advance the cursor at delivery, since their store write may never happen (abort) while their invalidation is already queued; their read-through directory pairs no store read with it, and rename fragments are applied first. Concurrent first opens raced: a slower opener's older lookup could overwrite the newer entry a faster opener had installed, while the monotonic version kept the newer timestamp — an old entry fenced at a new version, immune to every correcting event. Entry and version are now installed as one decision under the handle map lock, and an install that does not outrank the handle's version is dropped. The remote-cache commit also escaped the fence: it wrote storage and notified without the path lock, so a lookup's shared-locked fence and read could land between the two and hand out the cached state under-versioned. The commit now re-reads and writes under the exclusive path lock, and backs off entirely when the entry changed during the download — the concurrent writer supersedes the cached content. * mount: floors gate store applies; installs respect handle users; renames join the fence A directory floor certifies the listing state as of its snapshot, but a delayed event at or below the floor was still applied to the store — rolling the content back to pre-snapshot state while the floor kept claiming the snapshot version, so the correcting events were fenced out of every future read. Events are now gated against the affected directory's floor, each half of a rename independently. Fences are lower bounds: a listing or lookup can include a mutation whose event has not been delivered yet, and that event later passes every gate carrying state the handle already holds. Such a re-delivery now advances the version without destroying dirty pages or reinstalling the entry — invalidating local writes over a no-op was the real damage in every remaining under-fence window, including the unlocked listing snapshot, which no per-path lock can serialize. The concurrent-open install moved from the map lock to the handle lock every reader, writer, and invalidation synchronizes on, and rejects what cannot improve the handle: dirty state (local writes would be lost), unversioned lookup responses (they cannot outrank anything, and two zero-version opens must not overwrite each other), and anything not strictly newer. New handles are still fully initialized before the map exposes them. Renames committed metadata and emitted events with no path lock, so a lookup could read the renamed state under a fence preceding its events. Both rename handlers now hold the source and destination locks, ordered by path, across commit and notification; descendants of a renamed directory are not individually locked and rely on the no-op re-delivery handling above. * mount: per-entry store versions replace the cursor and directory floors The store's aggregate versions — a global subscription cursor and per-directory listing floors — were versions at coarser granularity than the values they described, and every over-claiming bug in this series traced to that gap: an aggregate vouching for state its source never saw. Each store entry now carries the filer log position of the write that produced it — the event that applied it, or the listing snapshot that inserted it, recorded in the store's key-value space under the same lock as the entry write. The store becomes what the handle already is: a last-writer-wins register with one rule, install only what outranks the current claim. The cursor, the floors, their advancement rules, the pairing ordering constraint, and the floor gating all collapse into that rule. Applies are gated per entry, each half of a rename independently; an unversioned local write clears the claim its content no longer proves; version records lingering after a bulk folder wipe cannot fence a recreate, since a claim only blocks while its entry exists. Listing inserts are stamped at build completion, before the buffered replay so newer replayed events override the stamp. Filer side, the fence dance every versioned read must perform is now a single choke point, fencedFindEntry, so a future read RPC gets the lock-serialized stamp by construction rather than by convention. * mount: judge no-op re-deliveries against an immutable base, not the live entry The equal-state skip compared the incoming event to the live handle entry, but local writes mutate the live entry — size, timestamps, chunks — so a delayed event re-delivering the base the handle was opened with no longer matched, and the installer destroyed the dirty pages and rolled the entry back over nothing new. The handle now keeps an immutable snapshot of the filer state it last installed or acknowledged, refreshed at every install and mutation ack (flush acks snapshot the request entry before the id mapping mutates it), and the no-op judgment runs against that base: an event carrying the base brings nothing, whatever the live entry has diverged to since. * mount: tombstones for versioned deletes, absence floors, copy enrollment Four gaps in the per-entry version protocol, all the same shape: a versioned fact with nothing carrying its version. A deletion is a fact about a path with no entry left to hold it — clearing the record let a delayed older event resurrect the deleted path, permanently, since the deletion's own redelivery is dedup-suppressed. Versioned deletes now leave a tombstone record that fences without an entry; renames tombstone their source the same way. Plain records still only block while their entry exists, so records lingering after a bulk folder wipe cannot fence a recreate. A completed listing proves absences as well as presences: a name it omitted was deleted as of the snapshot, and a delayed create below the snapshot re-creates it. The snapshot is kept per directory strictly as an absence fence, consulted only when a path has neither an entry nor a version record — present entries carry their own versions and never touch it, which is what separates this from the over-claiming floor it replaces. A rebuild against a pre-upgrade filer returns no snapshot; stamping now clears the children's records in that case, so a reinserted entry cannot reactivate the stale claim its previous incarnation left behind and reject valid events below it. Server-side copies installed the copied entry without enrolling in the base protocol, so the copy's own event differed from the stale pre-copy base and destroyed writes made to the destination after the copy. The install now refreshes the base and takes its version from the fenced readback. * mount: deletion facts outlive the cache's knowledge of the entry A versioned delete of a path the store held no entry for recorded nothing, so a delayed older event recreated the path — permanently, with the deletion's redelivery dedup-suppressed. The tombstone is now written whenever a versioned event vacates a path: the deletion is a fact about the path, not about what this cache happened to hold. For an absent entry, the listing's absence floor now speaks whatever older record remains: a tombstone at one position does not exhaust what is known about the path when a newer snapshot has confirmed the name still absent, and an event between the two was slipping past both. A committed copy whose readback failed installed a synthesized base with local timestamps; the copy's real event legitimately differs from it, and was read as foreign state — destroying writes made to the destination after the copy. The handle now marks that its own event is en route and adopts that event's state as the base without touching the live entry or the dirty pages; the adoption is one-shot, so a genuinely foreign event still invalidates. * mount: authoritative acks cancel pending event adoption; tombstones scoped and pruned The copy-event adoption flag could outlive its purpose: a flush after the failed readback installs a newer base and advances the version, the copy's own event is then version gated without consuming the flag, and the next genuinely foreign event was silently adopted — base advanced, live entry and dirty pages untouched — leaving the mount to later overwrite that remote change. Every local acknowledgment now installs its base through one helper that also cancels any pending adoption: the ack supersedes the mutation the adoption was waiting for. Tombstones were written for every versioned delete under the mount and survived directory eviction by design, growing LevelDB with historical deletions on delete-heavy mounts. They are now scoped to directories whose cached state the fence actually protects — an uncached parent never serves from the store nor applies the resurrecting insert — and a completed listing prunes the direct-child tombstones its absence floor supersedes, leaving only those above the snapshot. The store gains a key-prefix visitor for the sweep. * mount: acked saves install their value; trailer snapshots; direct-child prune range A version must never advance without its value. saveEntry stamped any open handle with the acknowledgment's version, but a handle opened while the save was in flight holds the pre-mutation entry — stamping it fenced out the events carrying the state it lacked, permanently, with the local apply performing no invalidation and the redelivery deduplicated. The acknowledged entry is now installed together with its version, through the same guarded install the racing-open path uses: under the handle lock, only when it outranks the handle, never over dirty local writes. Empty listings return no in-band snapshot — a snapshot-only response would be read as an entry by older consumers — so directories that end empty gained no absence floor and their tombstones were never pruned. The filer now sends the snapshot in the stream trailer, which older clients ignore, and the client reads it when no in-band snapshot arrived. Empty directories get real floors, their tombstones prune, and their buffered replays gain the snapshot filter instead of the replay-all fallback. Version records now encode the parent directory and name separated by a NUL, making a directory's direct children one contiguous key range: the tombstone prune scans exactly them under the cache lock, instead of walking every descendant record — the whole store, for root. * mount: fix dirty-page loss, uid/gid base, download race, copy adopt, leak; dedup Correctness fixes from the versioned-invalidation review: - A foreign delete/rename-away of a file held open with unflushed local writes destroyed the dirty pages unconditionally. A process may keep writing to an unlinked-but-open file and those writes were already acknowledged; preserve the pages when the handle is dirty. - downloadRemoteEntry stored the handle's base with filer-side uid/gid while every candidate it is later compared against is in local form, so under a non-identity UidGidMapper an unchanged re-delivery looked foreign and force-destroyed dirty pages. Map the base to local. - downloadRemoteEntry wrote the entry/base/version triple under only the handle's shared lock, so two concurrent reads of the same remote-only file could tear it. Serialize the install with a dedicated mutex (invalidation is already excluded by the exclusive handle lock). - A committed server-side copy whose readback failed adopted the FIRST event past the version gate as its base; a foreign write delivered first was silently swallowed. Adopt only an event whose content matches the synthesized base — the copy's own event — and install any other normally. - The deferred-create path relied on AcquireFileHandle installing the passed entry on a pre-existing handle, which the version rework dropped. Restore that install in the compat wrapper; the versioned open path keeps its gated install. Growth and hot-path cost: - Per-entry version records and tombstones leaked when a directory was evicted or read-through without a rebuild. An uncached directory gates its own inserts, so its records fence nothing; clear a directory's child version records when it is wiped for eviction. - FindEntry paid for the version KvGet on every lookup/getattr cache hit and threw it away. FindEntry now reads only the entry; the hot lookupEntry cache-hit path skips the version entirely. Cleanups: - Extract ackVersionTsNs over the shared response interface, replacing the metadata-event-else-log-ts snippet copy-pasted at four ack sites. - Extract acquireRenamePathLocks, replacing the verbatim sorted two-path lock fence in both rename handlers. * mount: no resurrection on foreign delete, version no-event acks, gate downloads, tighten copy adopt Follow-ups to the review patches: - Preserving dirty pages on a foreign delete let the next flush pass the isDeleted guard and CreateEntry, resurrecting the remotely-unlinked name. Mark the handle deleted in the vacate branch: the open fd can still read its buffered writes, but a flush no longer recreates the file. - A no-event acknowledgment (log fence only) synthesized a metadata event with TsNs 0, so the cache stored the entry unversioned and an older subscriber event rolled it back. Stamp the synthesized event with the ack's log position at all four ack sites. - downloadRemoteEntry serialized its install but did not check the version, so an older response arriving last overwrote the entry/base while the monotonic version kept the newer value, fencing corrections out. Install only when the response is at least as new as the handle. - sameEntryContent compared only size and chunks, so a foreign chmod with unchanged content was adopted as the copy's own event. Compare everything except server-assigned timestamps, so a metadata-only foreign change installs instead. * mount: trim comments to the non-obvious why The versioning work accumulated multi-line comment blocks restating what the code says. Keep the constraint a reader cannot derive — why a fence is exact, why a version must not advance without its value, why an uncached parent's records fence nothing — and drop the rest. * mount: distinguish rename from delete, tighten the download and adopt gates - A rename emits a nil old-path invalidation just like an unlink, so the vacate branch marked the handle deleted and later writes through the already-open descriptor were skipped instead of persisted. Carry the delete/rename distinction on the invalidation and mark only an actual delete. - The remote-download install accepted an unversioned response regardless of the handle's version, so during a rolling upgrade a delayed response could install stale content under a newer version. Require the response to be at least as new, with one exception: a handle still lacking local chunks takes the content anyway — it cannot read without it — but does not claim the response's log position. - Copy-event adoption returned without installing, so a foreign touch arriving before the copy's own event lost its timestamps. Content is unchanged either way, so the dirty pages stay valid; a clean handle now takes the entry, while a dirty one keeps its diverged version. * mount: one directory floor instead of a record per child; agree on TTL Review feedback: - Build completion wrote one KV record per direct child inside the cache write lock, so a large directory stalled every other cache operation for O(children) store writes. The directory's listing snapshot already covers every child it saw; make that floor the version for any child without a record of its own, and a child earns a record only when a later event touches it. One map write per build replaces the per-child writes, with the same fencing. - The presence probe read the store directly and so counted a TTL-expired entry as present, judging the path by a record describing content that has logically vanished. It now applies the same expiry the read path does, and an expired path falls back to its directory floor. - Preserve ErrNotFound identity when the commit-time re-read finds the object deleted, so callers still surface a 404. - Assert the rename-away source fence timestamp in the invalidation test. Also record the tombstone ceiling: distinct deleted names in a cached directory accumulate until it is rebuilt or evicted, which prunes everything at or below the new snapshot. * mount: pin the fence's clock domain instead of letting skew decide A log-position fence is stamped by one filer's clock under that filer's in-process lock, so comparing it to an event another filer logged is comparing two unrelated clocks. The two error directions are not equally costly: applying an event the fence already covered is a re-apply the base-equality check absorbs, while skipping one it does not cover leaves the handle holding exactly the state the event was meant to correct, with the subscription cursor already past it — the unhealable staleness this whole PR exists to remove. So refuse to guess. Fences now carry the signature of the filer that stamped them, and a handle records it alongside the position. An event is only fenced out when the filer that logged it is the one that stamped the fence — the logging filer appends its own signature, so its presence identifies the clock domain. Events from any other filer are applied. Positions taken from events keep comparing as before; the subscription already delivers those in order. The invalidation callback takes a struct now: it carries the path, entry, position, delete/rename distinction, and signatures, and was about to need a fifth positional parameter. * mount: follow a foreign rename; key page invalidation on content, not equality - A rename's old-path invalidation now carries the destination, and the handle follows the file there: an open fd tracks the inode, and leaving it on the old path made its next flush recreate that name instead of updating the renamed file. - Dirty pages overlay content, so only a content change invalidates them. Keying that on exact equality meant any timestamp-only event destroyed them, which the copy-adoption marker existed to paper over — a foreign touch could consume the marker and leave the copy's own event to drop the post-copy writes. Comparing content instead makes the marker unnecessary, so it is gone: a metadata-only event keeps the overlay, and a dirty handle keeps its diverged entry unless foreign content supersedes it. - A remote download response that is merely older is now refused even when the handle still lacks chunks; only an unversioned one is taken (and claims no position), since an older response's content predates what the handle reflects. - A refused or unversioned download no longer publishes to the metadata cache, where a zero-position event would clear the entry's version and let an older subscriber event roll the cache back. * mount: page invalidation keys on content alone; unversioned writes claim no position - sameEntryContent compared everything but timestamps, so a foreign chmod, chown, or xattr change counted as a content change and destroyed the dirty-page overlay. It was strict only to serve the copy-adoption marker, which is gone; its one caller now asks the question it actually needs — did the bytes change — so metadata-only events leave the overlay alone. - A rename over an existing file destroys that file, but its open handle was left live and still pointed at the name the renamed source now occupies, so its flush could overwrite it. MovePath already reports the displaced inode; mark that handle deleted. - An acknowledgment was refused whenever its position was numerically lower, even when a different filer stamped the fence it lost to. Two known, differing signatures mean unrelated clocks, so the comparison no longer applies there; unknown signatures still compare as before. - A local write with no log position behind it now records that explicitly instead of deleting its version record. Absence means the directory listing covers the path, which is why the snapshot floor applies; local content the listing never saw must not inherit it, or the events that would correct it are fenced out. * mount: widen the existing lookup functions instead of forking WithVersion twins The versioning work grew a parallel function for every accessor that needed to return a log position — lookupEntryWithVersion beside lookupEntry, maybeLoadEntryWithVersion beside maybeLoadEntry, FindEntryWithVersion beside FindEntry, AcquireFileHandleWithVersion beside AcquireFileHandle, advanceEntryVersion beside advanceEntryVersionTsNs, plus a getPbEntryWithVersion wrapper and an InsertListedEntriesForTest hook. Two names for one operation is two places to keep in step, and the split let callers pick the one that happened to compile. Each pair is now the single original name carrying the position, with callers that do not want it discarding it. filer_pb.GetEntry returns the fence its response already carried rather than a mount-side wrapper re-issuing the lookup, and InsertEntry takes the position its content reflects rather than a test-only twin that inserted without one. The one behavioural knot the merge exposed: AcquireFileHandle had been installing the entry on a pre-existing handle only in its unversioned form, which conflated 'the caller is authoritative' with 'the lookup had no version'. Deferred create is the only caller that means the former, so it now installs explicitly and the map function just acquires.	2026-07-23 17:44:02 -07:00
Chris Lu	495632730c	feat(s3/lifecycle): daily-replay observability — metrics + summary log (Phase 6) (#9462) ... * feat(s3/lifecycle): daily-replay observability metrics + per-run summary log Operators have no Prometheus signal today for the daily_replay path beyond the cluster-rate-limiter wait histogram. Phase 6 adds the three baseline questions: how long does a shard take, how many events did it scan, and what did dispatch produce. - S3LifecycleDailyRunShardDurationSeconds (histogram, label=shard): wall-clock per shard. p95 climbing toward MaxRuntime means the shard is brushing its budget. - S3LifecycleDailyRunEventsScanned (counter, label=shard): meta-log events drainShardEvents processed. Pairs with the duration so a spike in events-per-shard correlates with a slow shard. - S3LifecycleDispatchCounter (existing, reused): processMatches now increments this with the outcome label, so streaming and daily_replay paths share one outcome view. Transport errors are counted under outcome="TRANSPORT_ERROR". dailyrun.Run logs a per-run summary at V(0): status / shards / errors / duration. The summary is the at-a-glance line operators read in /var/log to confirm a run completed. Test pins the dispatch-counter increment with a unique bucket/kind/outcome triple so a refactor that drops the instrumentation call surfaces as a test failure. * fix(s3/lifecycle): align dispatch error label + clean test labels Two PR-9462 review fixes from gemini: 1. processMatches' transport-failure label was "TRANSPORT_ERROR"; streaming's dispatcher uses "RPC_ERROR" for the same condition (see dispatcher/dispatcher.go). Use "RPC_ERROR" here too so the same Prometheus query covers both delete paths. 2. The dispatch-counter assertion test now deletes its label row on exit so the in-process Prometheus registry doesn't accumulate per-test state across the suite.	2026-05-12 12:15:20 -07:00
Chris Lu	122ca7c020	feat(s3/lifecycle): daily-replay worker behind algorithm flag (Phase 2) (#9446) ... * docs(s3lifecycle): design for daily-replay worker Captures the algorithm and dev plan iterated on in PR #9431 and the discussion leading up to it: per-shard daily meta-log replay, walker as a per-day pass for ExpirationDate/ExpiredDeleteMarker/NewerNoncurrent plus a recovery branch over engine.RecoveryView(snap), explicit retention-window input to RulesForShard, two cursor hashes (ReplayContentHash + PromotedHash) that together detect every invalidation case. Implementation phases are sequenced so each can ship independently — Phase 1 (noncurrent_since stamp) just landed. * feat(s3/lifecycle): daily-replay worker behind algorithm flag (Phase 2) New weed/s3api/s3lifecycle/dailyrun package implementing the bounded daily meta-log scan from the design doc. One pass per Execute per shard: load cursor, scan events forward, route each through router.Route, dispatch any due Match, advance the cursor on success. Halt-on-failure keeps the cursor at the last fully-processed event so tomorrow resumes from the same point — head-of-line blocking is the deliberate failure signal. Replay-only in this phase. Phase 4 wires the walker for ExpirationDate, ExpiredDeleteMarker, NewerNoncurrent, and scan_only-promoted rules. Until then a typed UnsupportedRuleError refuses runs on those buckets: operators see the rejection in the activity log rather than silently losing rules. Behavior: - Per-shard cursor {TsNs, RuleSetHash, PromotedHash} JSON-persisted under /etc/s3/lifecycle/daily-cursors/. PromotedHash always-empty in Phase 2; Phase 4 turns it on. - Rule-change branch rewinds cursor to now - max_ttl when the replay-content hash mismatches. Cold start uses the same floor. - Transport errors retry 3x with exponential backoff capped at 5s; server outcomes (RETRY_LATER / BLOCKED) halt the run without retry. - Empty-replay sentinel: cursor TsNs=0 when no replay-eligible rules exist, only the hash gates a future addition. Worker shape: - New admin config field "algorithm" with enum streaming|daily_replay, default streaming. Existing deployments are unaffected. - handler.Execute branches on the flag: streaming routes through the current scheduler.Scheduler, daily_replay routes through dailyrun.Run. - dispatcher.NewFilerSiblingLister exported so both paths share the same .versions/ + null-bare lookup. Engine integration: - Local replayContentHash + maxEffectiveTTL helpers in dailyrun. Phase 4's engine surface (ReplayContentHash, MaxEffectiveTTL) will replace them with one-line redirects; the local versions hash the same fields so the cursor stays valid across the swap. Tests cover cursor persistence, unsupported-rule rejection, hash stability under rule reordering, hash sensitivity to TTL edits, max-TTL aggregation, dispatch retry budget, and request shape including the identity-CAS witness. Includes the design doc at weed/s3api/s3lifecycle/DESIGN.md so reviewers and future phases share the same spec. * feat(s3/lifecycle): default to daily_replay; streaming becomes the fallback knob The streaming dispatcher hasn't shipped to users yet, so there's no backward-compat surface to preserve. Flip the algorithm default from streaming to daily_replay so the new path is the standard from day one. Streaming stays as an explicit opt-in escape hatch during the Phase 4 walker rollout; Phase 5 deletes both the flag and the streaming code. Buckets whose lifecycle rules require walker-bound dispatch (ExpirationDate, ExpiredDeleteMarker, NewerNoncurrent, scan_only) will fail the daily_replay run with the existing UnsupportedRuleError until Phase 4 walker integration ships. Operators hitting that case can set algorithm=streaming until the follow-up lands. Updates the test for the default value and renames the unknown-value-fallback case to reflect the new default. * fix(s3/lifecycle/dailyrun): drop per-rule done flag — it suppressed due matches The done map was keyed by ActionKey = {Bucket, RuleHash, ActionKind}. That's only safe when each event produces at most one match per ActionKey with a single deterministic due-time formula — ExpirationDays and AbortMPU fit that shape because due_time = ev.TsNs + r.days is monotonic in event TsNs. But NoncurrentDays paired with NewerNoncurrentVersions > 0 (allowed in Phase 2 since it compiles to ActionKindNoncurrentDays) routes through routePointerTransitionExpand, which emits matches for every noncurrent sibling — each with its own SuccessorModTime taken from the demoting event for that specific sibling. A single event can therefore produce two matches for the same ActionKey on different objects with wildly different DueTimes. With the old code, a not-yet-due sibling encountered first would set done[ActionKey] = true and then the next sibling — even though its DueTime had already passed — would be skipped. Future events for the same rule would also be suppressed for the rest of the run. Objects that should have been deleted weren't. Fix: drop the early-stop optimization. Process every match independently. A future-DueTime match is now silently skipped without affecting any later match. The performance hit is small (Phase 2 is a single bounded daily pass, and the rate limiter is the real throughput governor); the correctness gain is non-negotiable. Also fixes the inverted comment in processMatches that described the old check as "due_time is past now" when it actually checked DueTime.After(now) (i.e., NOT yet due). Adds four targeted tests: - not-yet-due match first in slice does not suppress two later due matches for the same rule; - reversed slice ordering produces identical dispatch; - BLOCKED outcome halts the loop before later due matches are sent; - empty match slice is a no-op. Phase 4's walker-and-recovery integration can revisit a per-(rule, object) memoization if profiling argues for it. * fix(s3/lifecycle/dailyrun): address PR review — cursor advance, mode gate, ctx cancel, snapshot consistency Addresses PR #9446 review feedback. Eight distinct fixes: 1. CURSOR ADVANCEMENT (gemini, critical). The old code advanced the persisted cursor to lastOK = TsNs of the last event processed, including events whose matches were skipped as not-yet-due. Those skipped matches would never be re-scanned, so objects under long-TTL rules would never expire. Track a "stuck" flag in drainShardEvents: the first event with a skipped (future-DueTime) match stops cursorAdvanceTo from rising, but the loop keeps processing later events to dispatch any that ARE due. The persisted cursor sits at the last fully-processed event so tomorrow's run re-scans from the skipped event onward and the future-due matches get re-evaluated when they age in. processMatches now returns (skippedAny, halted, err) so the drain loop can tell apart "event fully drained" from "event had pending future-due matches." 2. MODE GATE (gemini). checkSnapshotForUnsupported only checked the ActionKind. A replay-eligible kind with Mode != ModeEventDriven (e.g. ModeScanOnly via retention promotion) passed the check but then got silently ignored by router.Route, which gates dispatch on Mode == ModeEventDriven. Reject loudly with the typed error so admin sees the rejection in the activity log. 3. WORKERS CONFIG (gemini). The handler hardcoded 16 concurrent shard goroutines regardless of cfg.Workers. Add a Workers field to dailyrun.Config and gate the goroutine fan-out on a semaphore of that size; the handler now passes cfg.Workers through. 4. SINGLE SNAPSHOT PER RUN (coderabbit). Run() validated against one snapshot but runShard() pulled a fresh cfg.Engine.Snapshot() per shard. Mid-run Compile would let shards process different rule sets. Capture snap at the top of Run, pass it down to every shard. 5. FROZEN runNow (coderabbit). drainShardEvents and processMatches accepted a `now func() time.Time` and called it multiple times. DueTime comparisons would slip as the run wore on. Capture runNow once at the top of Run and thread it through as a time.Time value. 6. CTX CANCELLATION (coderabbit). The drain loop's <-ctx.Done() case broke out of the loop and returned nil, marking interrupted runs as successful. Return ctx.Err() instead so the caller propagates the interrupt; cursorAdvanceTo carries whatever progress was made. 7. CURSOR LOAD VALIDATION (coderabbit + gemini). The persister silently accepted empty files, mismatched shard_ids, and hash slices shorter than 32 bytes (copy() would zero-pad). Each now returns a typed error so the run halts and an operator investigates rather than silently re-scanning from time zero or persisting a zero-padded hash that masks corruption forever. 8. DEAD BRANCH (coderabbit). The "lastOK < startTsNs → keep persisted" guard in runShard was unreachable because drainShardEvents initialized lastOK := startTsNs and only ever raised it. Removed along with the new cursor-advancement semantics that handle the "no events processed" case implicitly. Plus markdown lint: DESIGN.md fenced code blocks now carry a `text` language identifier to satisfy MD040. Skipped from the review: - gemini's "maxTTL == 0 incorrectly skips immediate expirations": actions with Days <= 0 don't compile to a CompiledAction (see weed/s3api/s3lifecycle/action_kind.go: `if rule.X > 0`). The new empty-replay sentinel uses `rsh == [32]byte{}` for clarity per gemini's suggested form, but the behavior is equivalent. Tests added/updated: - TestProcessMatches_AllDueNoSkippedFlag pins skippedAny=false when all matches are past their DueTime. - TestCheckSnapshotForUnsupported_NonEventDrivenModeRejected pins the new Mode check. - TestFilerCursorPersister_EmptyFileReturnsError, _ShardIDMismatchReturnsError, _HashLengthMismatchReturnsError pin the new validation rules. - Existing process-matches tests reshaped for the (skippedAny, halted, err) return tuple. Full build clean. Dailyrun + worker test packages green.	2026-05-11 18:07:17 -07:00