Closed pranavidandu closed 2 months ago
Do you suggest we set lock_timeout
only for the CREATE_REPLICATION_SLOT
command sent over the replication connection? What do you think is a meaningful value?
Can you share the next lines of output from pgcopydb. I believe the next lines of the logs after Failed to send CREATE_REPLICATION_SLOT command:
(notice the colon at the end) should contain more information on the error.
@pranavidandu Waiting for the logs as @hanefi mentioned
Hi @pranavidandu ,
After some more internal discussions with the team we decided to keep the current behavior. Any operator can set lock_timeout
setting in Postgres and see the errors sooner than the config value that is 5 hours in this example.
If you read my earlier message, you can see more details on the error message in our logs. PostgreSQL waits for locks, and reports that in their logs as well.
Migration failed to create snapshot due to concurrent transactions and failed after 5 hours. Can we have a timeout in these scenarios and inform the customer about this error?
Source server logs:
Pg stat sessions logs:
Repro link - https://issues.redhat.com/browse/DBZ-862