Conversation
…topics in recovery
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Closes #141.
SubscriptionCache::reconciletreated the local topic index as the source of truth, even though that index is filled only by a one-hop broadcast. When a node got the replicated subscription record but missed the broadcast, reconcile deleted the record,export_partitionpassed the loss on, and the node kept failing to route to that subscriber.clear_partitionhas no production caller). Without the restriction, an ex-owner turned its stale copy into routing entries for clients that had already unsubscribed. This was the quorum review's main finding; it is reproduced live below.resp/…) are skipped in reconcile and now also in startup recovery, since their subscribe and unsubscribe are never broadcast. The check is one sharedis_response_topicfunction, replacing four copies.WildcardPendingStoreand its 1-minute timer. Only its own tests calledadd_pending, so the timer never did anything.specs/ClusterSubReconcile.tlamodels holders P and R, a non-holder O, and a former holder S with a stale copy.handle_forwarded_publish_no_dedup). Theholderghostcfg records this.Test plan
cargo make clippycargo make testreplicated_subscription_survives_reconcile_when_broadcast_was_missedfails on main (left: []), passes here.reconcile_ignores_records_for_partitions_not_heldreconcile_skips_response_topics(now also checks the record keeps the entry)recovery_rebuilds_index_without_response_topics(fails before the recovery change)InvNoStaleCopyGhostandRoutingRepaired(exhaustive at MaxOps=2);unrestrictedviolatesInvNoStaleCopyGhostat depth 6;holderghostviolatesInvNoHolderGhost(accepted, see above).oldlossfails at depth 4,oldresurrectat depth 8;norepairandnonholderfail their liveness properties, as expected.removed=1reconcile on Q per run.