On 01/14/2014 04:55 PM, Brian Fehrle wrote:
> On 01/06/2014 12:43 PM, Scott Marlowe wrote:
>> On Mon, Jan 6, 2014 at 12:15 PM, Jan Wieck <JanWieck at yahoo.com> wrote:
>>> On 01/06/14 13:51, Brian Fehrle wrote:
>>>> ...
>>>>
>>>> 2014-01-06 11:43:22 MST DEBUG1 remoteHelperThread_1_1: 112.859 seconds
>>>> delay for first row
>>>> 2014-01-06 11:43:22 MST DEBUG1 remoteHelperThread_1_1: 112.906 seconds
>>>> until close cursor
>>>> 2014-01-06 11:43:22 MST DEBUG1 remoteHelperThread_1_1: inserts=61
>>>> updates=300 deletes=55 truncates=0
>>> Almost 2 minutes to select 416 log rows? That looks wrong.
>>>
>>> Can you change the slon configuration to have an explain_interval? With
>>> that slon will emit EXPLAIN output for the log selection query in that
>>> interval. Maybe something is screwing up the optimizer.
>> I always check log_1 and log_2 for bloat when these things start
>> happening. Occasionally it's seqlog having a problem.
> I don't see a ton of bloat on either the master or the slave. There was
> some on the slave but after a vacuum full (and reindex) we see no
> speedup in replication.
>
>
> Right now I'm trying to see if I can see exactly what the events in
> sl_log are for a given event ID. How do I tie these together? I get the
> latest sync event from the view sl_status, and can see some basic info
> in sl_event, but what in there helps me tie it to the rows in
> sl_log_[1-2] so I can get an idea of just what the events are, in case
> it's something crazy big.
>


It isn't easy to do manually we don't have a stand-alone script that 
does this.  You need to look at the snapshot of the last SYNC that was 
applied, and the snapshot of the next sync your applying.

Any xid's that had not committed yet in the last SYNC (because they were 
in progress or had not yet started) but were no longer in progress in 
the next SYNC need to be applied.

You then select any rows from sl_log_1 and sl_log_2 that have xid's in 
the above list.  The code for this is in sync_event in remote_worker.c 
look for the big queries with unions

Steve


> - Brian F
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general
>