On 7/25/2018 10:28 AM, Andres Freund wrote:
Are you really expecting us to be able to reproduce the problem based on the above description? Our test suites do setup plain replications setups, and the problem doesn't occur there.
I don't, by definition, have a reproducible case: it only happened once so far.
If nobody knows what limits the number of files created in $PGDATA/pg_logical/snapshots then we'll all have to wait until this happens again.
(To somebody else as I'm obviously not turning logical replication back on until I know it won't kill my server again.)
Dima