For some specific tables in RDS Aurora postgresql 16.2 - ERROR: invalid snapshot identifier

[debugger22]

[pg_query_executor] failed to set snapshot: ERROR: invalid snapshot identifier: "0000002F-00040D68-1" (SQLSTATE 22023)

[debugger22]

It's happening with large tables

[debugger22]

Sometimes it also throws - failed to set transaction snapshot: failed to set transaction snapshot: ERROR: invalid snapshot identifier: "0000000E-0002EEB6-1" (SQLSTATE 22023)

[serprex]

I've seen this happen when the initial copy gets interrupted. In that scenario we should get a new snapshot

Is peerdb running out of memory on your setup?

[Amogh-Bharadwaj]

Hi @debugger22 !
Could happen if peerdb - specifically the flow-snapshot-worker container - restarts. That container needs to always be up during initial load. Is it possible to check if that container ever went down?

[pmandalapu]

We are seeing similar error while replicating from Postgres to Snowflake, we did the inital sync and it been running in CDC mode. The replication stopped about a week ago with this error message

failed to pull records: [pg_query_executor] failed to set snapshot: ERROR: invalid snapshot identifier: "0000000E-00128E0A-1" (SQLSTATE 22023)

[heavycrystal]

that error can only occur in initial load, we'd need logs in the surrounding context

[nitindhiman314e]

Hi All, Did we get any workaround to solve this problem? I'm also getting below same error

failed to pull records: [pg_query_executor] failed to set snapshot: ERROR: invalid snapshot identifier: "000000E3-000062E4-1" (SQLSTATE 22023)

[serprex]

#2168

[nitindhiman314e]

#2168

Hi @serprex ,
Thank you for your response! I noticed that issue #2168 is still in an open state. Are you suggesting that the "fix-initial-snapshot-replay" branch contains the fix for the "invalid snapshot identifier" issue, or were you referring to something else?
Looking forward to your clarification!

[serprex]

That PR addresses this issue. It's held up being merged because there's worry about getting stuck in an error loop with initial load causing too much load or something