Julian Scarfe wrote:
> So with 1.2RC3 I'm still seeing the same behaviour as I reported below (no
> replies received to that).
>
> I have 10 sets (numbered 1001, 2001,... 10001) with initial origin on node
> 1.
>
> On attempting to switchover all of them to node 2, I'm getting FATAL errors
> on the 2nd and subsequent MOVE SETs:
>
> $ grep -a "MOVE" /tmp/slon-avbrief.out
> 2006-08-17 10:01:42 UTC DEBUG2 localListenThread: Received event 1,214
> MOVE_SET
> 2006-08-17 10:01:44 UTC DEBUG2 localListenThread: Received event 1,217
> MOVE_SET
> 2006-08-17 10:01:44 UTC FATAL localListenThread: MOVE_SET but no provider
> found for set 2001
> 2006-08-17 10:01:56 UTC DEBUG2 localListenThread: Received event 1,219
> MOVE_SET
> 2006-08-17 10:01:56 UTC FATAL localListenThread: MOVE_SET but no provider
> found for set 3001
> ...
> 2006-08-17 10:03:20 UTC DEBUG2 localListenThread: Received event 1,240
> MOVE_SET
> 2006-08-17 10:03:20 UTC FATAL localListenThread: MOVE_SET but no provider
> found for set 10001
>
> Of course with the new behaviour of slon in 1.2, it restarts after 10
> seconds and processes another MOVE SET successfully before dying at the
> second attempt. So eventually the process completes. As a side issue, if I
> could modify the sleep time before restart to 1 second, the FATAL might be
> acceptable, but 10 seconds per set is too long.
>
> Unless I'm misunderstanding, it looks like Christopher diagnosed the problem
> in the message referenced below, but I can't see any corresponding
> modification in HEAD to rewrite the query to include the set_id (sub_set).
>
> slon_mkquery(&query2,
> "select sub_provider
> from %s.sl_subscribe "
> " where
> sub_receiver = %d",
> rtcfg_namespace,
> rtcfg_nodeid);
> res2 = PQexec(dbconn, dstring_data(&query2));
> ...
> if (PQntuples(res2) != 1)
> {
> slon_log(SLON_FATAL,
> "localListenThread: MOVE_SET "
> "but no provider found
> for set %d\n",
> set_id);
> dstring_free(&query2);
> PQclear(res2);
> slon_retry();
> }
>
> Am I missing something?
>
I probably am...
I'll see to adding a test for this today (e.g. - a MOVE SET test with
~10 sets).
Adding "and sub_set = %d' to the query looks likely to work out OK...
_______________________________________________
Slony1-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/slony1-general
