Thread: Postgres dies on standby server after triggering failover

Hi.

I posted this question at the end of October, and didn't get any responses, so I'm gonna try one more time.

I have a problem with a standby server running on Windows 2003 R2, Enterprise x64 edition. I use Postgres 8.3 (installed to C:\Program Files (x86)\ )
Everything was working fine (base backup, archiving, recovery), until I wanted to test failover.
I created the trigger file, the restore_command returned a nonzero exit code, but postgres said "could not open file... No such file or directory", and died.
I'm not sure if this has anything to do with running a 32 bit postgres on a 64 bit machine, or whether it's a bug. Maybe postgres didn't think that the restore command failed?? As the restore_command, I tried using a compiled perl script, pg_standby, and even the system copy command. They all give the same result.
I should mention that whenever I start postgres (in recovery mode) from the service console, I get a message saying that postgres started and then stopped. I figured postgres was in some pending mode, waiting for the recovery to end before it can really start (accepting connections). Should I start it another way (like pg_ctl) when I'm doing recovery?

Since this is my first time implementing database replication I really can't tell what went wrong.
Any help would be greatly appreciated!
Ori

----------

Some dumps.
Postgres restores files, but gives an error after I created the trigger file, and the restore_command returned a nonzero error code:


Now postgres seemed to be stuck, or dead. Using Process Explorer I saw that one of the postgres.exe processes was running drwtsn32.exe (Dr Watson Postmortem debugger) indefinitely. When I kill drwtsn32, postgres dies too, resulting in the following lines in the log (nothing had been logged between 16:58:45 and 17:20:27)


When I looked at the Event Log, I saw an Application Error: