fix pg14 recovery fix#5
Merged
Merged
Conversation
Pulls in two fixes now merged upstream: - Gate the pg_publication_rel.prattrs column-list lookup on PG15+, fixing schema fetch on PG14 where the column does not exist. - Skip permanently unrecoverable tables during moonlink recovery instead of crash-looping the bgworker (e.g. after DROP DATABASE on a database that still contained mooncake tables). Cargo.lock picks up the new tracing dependency of moonlink_backend.
propagate_compatible_mode read ivorysql.compatible_mode with Spi::get_one, which pgrx routes through SpiClient::update. Its mark_mutable call invokes GetCurrentTransactionId, assigning the calling backend a real write transaction id as a side effect. The moonlink RPC issued later in create_table makes the walsender run CREATE_REPLICATION_SLOT ... USE_SNAPSHOT, whose snapshot builder waits for every in-progress write transaction to finish - including ours - while our backend sits in block_on waiting for that same RPC to return. The resulting circular wait hangs create_table forever and is invisible to the deadlock detector, since one side waits on a socket rather than a lock. Read the GUC through Spi::connect + client.select instead, which never marks the transaction mutable, so the backend keeps its virtual xid and the snapshot builder has nothing to wait for.
The moonlink bgworker binds its Unix socket asynchronously after postmaster start, so the first mooncake call in a fresh cluster could hit ECONNREFUSED and panic. Worse, the panic fired inside the STREAM LazyLock initializer, poisoning it and permanently breaking every subsequent mooncake call in that backend. Retry the connect with exponential backoff (50ms doubling, capped at 1s) for up to 10 seconds before giving up.
The extension-drop watcher decided whether pg_mooncake itself was being dropped with SELECT true INTO is_mooncake_drop ... LIMIT 1. When the query returns zero rows, SELECT INTO overwrites the variable with NULL even if it was initialized to false, and IF NOT NULL is not taken - so the cleanup body ran on every sql_drop event. Any DROP TABLE would then try to drop the moonlink replication slot and publication, and the watcher self-removed, leaving nothing to clean up on a real DROP EXTENSION. Use IF NOT EXISTS (...) instead, which has no NULL case.
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
2 participants
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.
No description provided.