AWS RDS support about pgreplay HOT 3 CLOSED

I need %c in log_line_prefix so that I can identify sessions.

Is it an option to use log_destination = csvlog? That would be better anyway, and the documentation seems to suggest that log_destination is modifiable.

from pgreplay.

Session id comes from process id and process start time. Documentation:

The %c escape prints a quasi-unique session identifier, consisting of two 4-byte hexadecimal numbers (without leading zeros) separated by a dot. The numbers are the process start time and the process ID

Could process id (%p) be made to substitute for session id?

I'll try log_destination = csvlog for future logs. In the meantime, I'm stuck with a pile of logs from just before a failure that are in the wrong format. I'd love to be able to replay those logs and see if I can reproduce the failure.

from pgreplay.

The process ID is not unique, it can be reused, so I cannot use it as a session identifier.

I understand your situation, because I was in exactly the same situation once. That made me write pgreplay!

What you could do is to pre-process the logs with sed or something similar to get them into a format that pgreplay can understand. The important thing is that you have log_min_duration_statement set to 0 (and not log_statement = 'all', which would log the statements after they are done).

You'd have to fix the timestamp to contain milliseconds, and you'd have to fake the session ID, maybe by using a zero timestamp part and the process ID in hex. The hard part are continuation lines if you have statements with line breaks in them. Each such continuation line will start with a tab, so don't modify such lines.

from pgreplay.