On 01/23/2015 06:31 AM, Petr Novak wrote:
> Hi all,
>
> I'd like to ask for help clarifying an issue I'm having.
>
> I've recently prepared new servers in another datacenter for some of our
> databases which I want to set up as a streaming replicas. There are
> several instances(clusters) with size ranging from 50-150GB. Some of
> them were set up with no issue. Three of them failed to start after
> pg_basebackup completed with:
>
> FATAL: could not access status of transaction 923709700
> DETAIL: Could not read from file "pg_clog/0370" at offset 237568: Success.
>
> (the clog file differed in each case of course..)
>
> As for PG versions one is 9.1.14 (on both master and replica) and the
> other two 9.2.9 (also on both)
>
> I've checked each file in question on the master server and in each case
> it was last changed during the run of the pg_basebackup. I've copied the
> file from master to the slave and it started normally, succesfully
> connected to primary and everything seems ok since then.
>
> My question is, why this is happening? Is it because the clog file was
> copied by pg_basebackup before the trasaction status had been written to
> it? Couldn't this information be recovered from xlog files?
> Is my current solution safe for the database (I suppose so, but I'd
> rather have it confirmed :) )
>
> I'm puzzled as I haven't found this issue discussed at all so far and it
> happened to me three times in one day on three different servers..
Not sure what is going on, but it would help to see the pg_basebackup
command issued.
>
> Thanks for any help.
>
> Cheers.
> Petr
>
>
--
Adrian Klaver
adrian.klaver@aklaver.com
