Marco Slot

There was a bug with this exact symptom that was fixed in October https://github.com/citusdata/pg_cron/blob/main/CHANGELOG.md I'm guessing they should update pg_cron

As long as the cron.jobs table is local to a specific node, then jobs would only run on that node. That's the case in Citus at least.

Could there be another shared_preload_libraries somewhere in your postgresql.conf or postgresql.auto.conf that overrides the first?

> How large can this get before thing go bad? Things go bad when 4 billion runs are pending, which can happen after 7605 years assuming jobs that run every...

> At a minimum, all code that specifies or specified a schema for pg_cron is now broken. In our case it's Terraform code. I think that's fair. Perhaps we should...

> Makes transaction recovery and distributed deadlocks detection respect citus.max_shared_pool_size I think this is probably the more controversial / hazardous part of this PR. It is required for stability that...

> I see your point, but it is still a more predictable way to manage a cluster, Perhaps, if we were starting from scratch. In practice, due to a large...

> Make citus.max_shared_pool_size cluster-wide. Per-database setting then may be used to set a connection quota for a database within the global citus.max_shared_pool_size Seems reasonable, not sure why we decided to...

> I assume, the way to solve this is to have one "management" database exactly for such cluster-wide procedures, and that is being implemented now in the pool_2pc branch? Yes,...

I wonder whether statistics gathering would be affected. It might anyway be good to do an ANALYZE as part of a move to get statistics.