Hi Álvaro,
Pardon my stupidity. I kept looking for things in the wrong place. The
pg_toast tables have always been there, I was just issuing the commands yo
mentioned in a different database. Need to get more sleep. xD
In the end, we decided the entire database needed a good cleaning. So we
scheduled a four-hour downtime, during which we restarted the database on a
different port (so that nothing would connect to it), increased the number
of autovacuum processes from three to five, and used vaccumdb to vacuum
analyze the top four dabatases in terms of usage.
We thought the pg_toast tables were still somehow corrupted, but a pg_dump
showed that we could dump it just fine. So we let them be. The next day,
performance was back to normal, and autovacuum was able to actually finish
working on these pg_toast tables by themselves. :")
Thank you so much for your help! I have questions about monitoring
autovacuum monitoring, but I'll create another thread for that!
Regards,
Tong
--
View this message in context:
http://postgresql.nabble.com/postgresql-9-3-5-autovacuums-stuck-on-non-existent-pg-toast-tables-tp5839397p5839897.html
Sent from the PostgreSQL - general mailing list archive at Nabble.com.