[Pgpool-general] pgpool-II working but not balancing load?

[Rolf Dybdal]

On Tue, Mar 1, 2011 at 7:41 PM, Tom Spivey <tom at spivey.de> wrote:

> Rolf Dybdal <rolf.dybdal at ...> writes:
>
> > > Hi everyone!My setup is; two postgres databases (9.0.3) replicated
> using
> streaming replication and one frontend running pgpool-II 3.0.3.The
> replication
> works great, and if I connect to the databases indvidually, I get the same
> result from my queries for both databases. > Pgpool also seems to work, and
> both
> db-servers get the expected number of connections set up, but when I check
> the
> traffic to the servers, I see that only one of the servers get the queries,
> and
> that the connections to the other server is just left idle. > If I detach
> (pcp_detach_node) the server that receives all the requests, the other one
> takes
> over. And when I attach the node, all queries are directed to that server
> again.
> In itself, a positive result, but not balanced. > Even if I set weight to 0
> on
> node 0 and 1.0 on node 1, node 0 still gets all traffic - and all queries
> are
> select only!I'm not sure if this behaviour is expected, I have tried
> generating
> traffic using JMeter, but wether I run with few threads or many, the
> situation
> is the same. > Log files show now qlues as far as I can see. I have deleted
> the
> pgpool_status file before restart (also tried -D param
> on startup).Anyone seeing the same?Best regardRolf Dybdal
> >
> > _______________________________________________
> > Pgpool-general mailing list
> > Pgpool-general at ...
> > http://pgfoundry.org/mailman/listinfo/pgpool-general
> >
>
> As you get both servers connected, the only obvious setting is:
> load_balance_mode = on
>
> The postgres logs should show the connections, extending the logs with the
> queries
> helps.
> You should find continous queries for the xlog situation on both servers.
>
> master: SELECT pg_last_xlog_receive_location()
> hot standby: SELECT pg_current_xlog_location()
>
> Depending on the settings, if the hot standby is too far behind, it will
> not get
> queries.
>
> This setting will show, if it is the case:
> log_standby_delay = 'always'
>
> That works for me, but I'm having other problems - still with failing
> queries...
>
> Cheers,
> Tom
>
> ---
Thanks Tom,
I have checked my settings, and they are indeed as you point out.
According to the postgres logs, databases are in sync, but I get an error
when i run pg_current_xlog_location() on the standby.
the error is; "WAL control functions cannot be executed during recovery"
which makes me wonder if the databases are NOT in sync after all.


Looks like I may have to start looking at the replication process, even if
that looks and works ok in all other tests.

Have anyone gotten and solved the "WAL control functions cannot be executed
during recovery" issue?

Thanks
Rolf


_______________________________________________
> Pgpool-general mailing list
> Pgpool-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-general
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pgfoundry.org/pipermail/pgpool-general/attachments/20110302/b6c14438/attachment.html>