Hi,
I am actually doing some test with pglogical and a trivial bi-directional replication setup (no BDR, just cross subscriptions, no conflicts because UUID a prim key).
During the tests I am stressing the autovacuum results to simulate a customer situation having trouble with files growing way different on two replica nodes.
I am checking for the count of dead tuples with pg_stat_user_tables or the extension pgstattuple and I am not sure if any is somehow accurate.
What is the most accurate way to check for this information except using pageinspect and/or freespacemap and count for myself and
what exact value is using by the autovacuum process related to the configured thresholds?
In another question few weeks ago I asked for the expected result using pglogical and autovacuum on both sides. It should be most likely the same
file sizes. But at the customers environment one site is always growing and there are just very small and short transactions.
Thanks and best regards
Dirk