fix: call pg_notification_queue_usage() automatically when the LISTEN channel bug is detected (#4858)

There's a PostgreSQL bug that doesn't let any listener to register in the DB: https://www.postgresql.org/message-id/flat/CAK98qZ3wZLE-RZJN_Y%2BTFjiTRPPFPBwNBpBi5K5CU8hUHkzDpw%40mail.gmail.com The only workaround is to advance the async notification queue tail, which can be done by executing: "SELECT pg_notification_queue_usage();". Before we just logged a HINT with this suggestion, but now we call that function directly and then let the listener to automatically recover. No automated tests were added here as it would be too complex and this is a PostgreSQL bug. But this was manually tested following the steps on https://github.com/PostgREST/postgrest/pull/4581#issuecomment-3690610592
2026-04-30 18:22:59 -05:00
parent 04a0e041e4
commit c09394517d
4 changed files with 16 additions and 8 deletions
@@ -22,6 +22,8 @@ All notable changes to this project will be documented in this file. From versio
 - Remove automatic transaction retries on `40001 (serialization_failure)` errors to prevent replication lag by @laurenceisla in #3673
 - Fix unexpected results when embedding and filtering the same table more than once by @laurenceisla in #4075
 - Restore Listener query shape so it can be found in pg_stat_activity by @mkleczek in #4857 #4859
+- The LISTEN channel now automatically recovers when it stops working due to a PostgreSQL bug @laurenceisla in #3147
+

 ### Changed