Conversation

Index corruption shouldnt happen... But life can be messy: okt.to/iAhu1s Gonna be discussing this live at the top of the hour: okt.to/DMX2sl

enterprisedb.com

Index Corruption in PostgreSQL: The Hidden Cost of Your Queries

A corruption in the database base often manifests itself in the form of errors when performing a DDL or DML operation on affected relations. The errors often show up as errors reading tables due...

Replying to

We see Postgres index corruption a fair bit at Discourse. Pg10 but switching to pg12 very soon.

Replying to

and

I work on the Postgres B-Tree code. Corruption simply should not happen with any regularity. E-mail me (address is pg@bowt.ie), or send a report to one of the community lists. Always interested in hearing problem reports, even when they do not involve bugs.

10:41 PM · May 23, 2020Twitter Web App

Replying to

and

Thank you! We're currently upgrading to postgres 12 and we'll see how it goes. At

@discourse

we upgrade every other version, so we are on version 10 and skipped 11. cc

Replying to

Cool! There are a bunch of B-Tree space utilization improvements in 12 that I worked on, but they're not expected to affect correctness or reliability. I strongly encourage root cause analysis of index corruption. It shouldn't be something you have to tolerate or accept.

Show replies