Read replica lag nearly filled Photoroom's primary database disk during a backfill
During a large backfill, a lagging read replica at Photoroom caused PostgreSQL to retain write-ahead logs on the primary, growing them from about 5 GB to 31.7 TB. Automatic storage growth had pushed provisioned disk to around 43 TB, closing in on a hard 60 TB limit. The team estimated only five to seven days remained before exhaustion, while the backfill needed several more weeks. They paused the work and enlarged the replica, which let the logs drain and the database return to normal. The team now monitors retained WAL alongside replica lag and limits backfill concurrency and transaction size.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in