Re: Warm standby recovery failure

Lee Azzarello <lee(at)dropio(dot)com> writes:
> cp: writing `pg_xlog/./000000010000002F000000AA': No space left on device
> 2009-01-29 12:48:14 UTC LOG: could not read from log file 47, segment
> 170, offset 3129344: Success
> 2009-01-29 12:48:14 UTC LOG: redo done at 2F/AA2FBE08

The bottom line here seems to be that your logfile transfer process
passed a damaged (partial) WAL file to the slave postmaster, which
interpreted the damage as being end of WAL. I'm not sure whether we
could find a way to prevent it from doing that without causing bigger
problems. In any case the most useful solution is going to involve
fixing the transfer script so that it notices failures from 'cp' and
doesn't pass the broken file on to the slave.

We probably should add a caution about this to the manual's discussion
of how to write archiving scripts.

regards, tom lane