Always update VM at replica

Don't split WAL record across two XLogData's when sending from safekeepers.
As protocol demands. Not following this makes standby complain about corrupted WAL in various ways. https://neondb.slack.com/archives/C05L7D1JAUS/p1703774799114719 closes https://github.com/neondatabase/cloud/issues/9057
2026-05-15 20:20:38 +00:00 · 2024-01-02 11:01:06 +02:00 · 2024-01-02 10:50:20 +04:00 · 2024-01-02 10:50:20 +04:00 · 2024-01-02 10:50:20 +04:00 · 2024-01-02 10:50:20 +04:00
1 changed files with 7 additions and 0 deletions
--- a/pgxn/neon/pagestore_smgr.c
+++ b/pgxn/neon/pagestore_smgr.c
@@ -2907,6 +2907,13 @@ neon_redo_read_buffer_filter(XLogReaderState *record, uint8 block_id)
 	if (!OidIsValid(NInfoGetDbOid(rinfo)))
 		return false;

+	/*
+	 * Always apply updates of VM because otherwise index-only scan may return wrong results
+	 * based on the wrong assumption that page is still all visible
+	 */
+	if (forknum == VISIBILITYMAP_FORKNUM)
+		return false;
+
 	CopyNRelFileInfoToBufTag(tag, rinfo);
 	tag.forkNum = forknum;
 	tag.blockNum = blkno;
Author	SHA1	Message	Date
Konstantin Knizhnik	e2886d55bc	Always update VM at replica	2024-01-02 11:01:06 +02:00
Arseny Sher	e92c9f42c0	Don't split WAL record across two XLogData's when sending from safekeepers. As protocol demands. Not following this makes standby complain about corrupted WAL in various ways. https://neondb.slack.com/archives/C05L7D1JAUS/p1703774799114719 closes https://github.com/neondatabase/cloud/issues/9057	2024-01-02 10:50:20 +04:00
Arseny Sher	aaaa39d9f5	Add large insertion and slow WAL sending to test_hot_standby. To exercise MAX_SEND_SIZE sending from safekeeper; we've had a bug with WAL records torn across several XLogData messages. Add failpoint to safekeeper to slow down sending. Also check for corrupted WAL complains in standby log. Make the test a bit simpler in passing, e.g. we don't need explicit commits as autocommit is enabled by default. https://neondb.slack.com/archives/C05L7D1JAUS/p1703774799114719 https://github.com/neondatabase/cloud/issues/9057	2024-01-02 10:50:20 +04:00
Arseny Sher	e79a19339c	Add failpoint support to safekeeper. Just a copy paste from pageserver.	2024-01-02 10:50:20 +04:00
Arseny Sher	dbd36e40dc	Move failpoint support code to utils. To enable them in safekeeper as well.	2024-01-02 10:50:20 +04:00
Arseny Sher	90ef48aab8	Fix safekeeper START_REPLICATION (term=n). It was giving WAL only up to commit_lsn instead of flush_lsn, so recovery of uncommitted WAL since `cdb08f03` hanged. Add test for this.	2024-01-01 20:44:05 +04:00
Arseny Sher	9a43c04a19	compute_ctl: kill postgres and sync-safekeeprs on exit. Otherwise they are left orphaned when compute_ctl is terminated with a signal. It was invisible most of the time because normally neon_local or k8s kills postgres directly and then compute_ctl finishes gracefully. However, in some tests compute_ctl gets stuck waiting for sync-safekeepers which intentionally never ends because safekeepers are offline, and we want to stop compute_ctl without leaving orphanes behind. This is a quite rough approach which doesn't wait for children termination. A better way would be to convert compute_ctl to async which would make waiting easy.	2024-01-01 20:44:05 +04:00