On 4/4/16 10:57 PM, Michael Paquier wrote:
> On Sun, Apr 3, 2016 at 2:50 PM, David Caldwell <david@porkrind.org> wrote:
>> Hello,
>>
>> We're using streaming replication. Our technique for spinning up a db
>> slave is this:
>>
>> rsync from master (gross copy)
>> pg_start_backup() on server
>> rsync from master (correct copy)
>> pg_stop_backup()
>> drop in recovery.conf into slave directory
>> enable hot_standby in slave conf
>> start slave
>>
>> After starting the slave, I'm getting this error every 5 seconds in the log:
>>
>> ERROR: requested WAL segment 0000000100000E2200000005 has already been
>> removed
>
> Are you sure that this is from the slave and not another not
> requesting it? If this log entry has just ERROR, it means that the WAL
> sender generated it. So don't you have a cascading slave connecting to
> this new slave and requested for this WAL segment that has already
> been removed?
Aha! Yes! There was a secondary slave still trying to replicate from the
original master (now turned into a slave).
> Should the error come from the WAL receiver, you would have something
like that.
> FATAL: could not receive data from WAL stream: ERROR: requested WAL
> segment 00000001000000000000001C has already been removed
I see the difference in the error messages now and it all makes sense.
Turning off that secondary slave causes the messages to stop. So it
looks like everything is good.
Thanks!
-David