[Pgcluster-general] PGCluster 1.7 seems to be stuck in deadlocks

[Hari G]

At.Mitani -

I have more information on the problem. It seems that PGCluster works fine with our Java application, polaris. It seems like the selects from this app is not replicated by PGCluster. However the geoserver app (another Java application) encounters many deadlocks as soon as it starts. IMHO, Geoserver is doing selects in a way that makes PGCluster replicate those. As soon as I start the Geoserver threads, I see locks forming in both the clusters. The second cluster's locks are getting released after few seconds but the locks in the primary cluster seems to be stuck forever. Why is this happening? I am sure we are only doing just read - no updates till now.

Any ideas?

Regards,
-- Hari


From: hari_g at hotmail.com
To: pgcluster-general at pgfoundry.org
Date: Thu, 27 Dec 2007 22:58:54 -0800
Subject: Re: [Pgcluster-general] PGCluster 1.7 seems to be stuck in deadlocks







At.Mitani -

Thanks for the quick reply. I have already tried setting the not_replicate_prepared_select to 'true' - now I tried with 'yes' also. That did not make any difference. With just one user and that too after using couple of minutes, I see 108 entries in pg_locks... which seem to be too much.
dikpala-cosme=# select count(*) from pg_locks;
 count 
-------
   108
(1 row)


Why did you call the variable "not_replicate_prepared_select" - is it only applicable for prepared select and not for dynamic selects?

Anything else?

Regards,
Hari Gangadharan

The /etc/hosts is as follows (also the ifconfig to show ipaddress):
# cat /etc/hosts
127.0.0.1       localhost node-1
10.252.71.16   cluster_1 rep_1 lb_1
10.252.47.70   cluster_2 rep_2 lb_2 db-host

# ifconfig
eth0      Link encap:Ethernet  HWaddr 12:31:38:00:28:B8  
          inet addr:10.252.47.70  Bcast:10.252.47.255  Mask:255.255.254.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:2908807 errors:0 dropped:0 overruns:0 frame:0
          TX packets:3154803 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000 
          RX bytes:662880498 (632.1 MiB)  TX bytes:1652103126 (1.5 GiB)
....


# cat /etc/hosts
127.0.0.1       localhost node-1
10.252.71.16   cluster_1 rep_1 lb_1
10.252.47.70   cluster_2 rep_2 lb_2 db-host
# ifconfig
eth0      Link encap:Ethernet  HWaddr 12:31:38:00:40:E2  
          inet addr:10.252.71.16  Bcast:10.252.71.255  Mask:255.255.254.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:1119951 errors:0 dropped:0 overruns:0 frame:0
          TX packets:883258 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000 
          RX bytes:254151981 (242.3 MiB)  TX bytes:243019474 (231.7 MiB)
...


> Date: Fri, 28 Dec 2007 09:25:00 +0900
> From: mitani at sraw.co.jp
> To: pgcluster-general at pgfoundry.org
> Subject: Re: [Pgcluster-general] PGCluster 1.7 seems to be stuck in deadlocks
> 
> Hi Hari,
> 
> Would you let us know the "/etc/hosts" in each servers.
> Did you set "yes" to "not_replicate_prepared_select" in the postgresql.conf ?
> If you don't use stored procedure with SELECT query, I recomend to set "yes" for it.
> Otherwise all queries are replicated when application uses JDBC.
> 
> Regards,
> ----------------
> At.Mitani
> 
> _______________________________________________
> Pgcluster-general mailing list
> Pgcluster-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgcluster-general

Don't get caught with egg on your face. Play Chicktionary! Check it out!

_________________________________________________________________
i’m is proud to present Cause Effect, a series about real people making a difference.
http://im.live.com/Messenger/IM/MTV/?source=text_Cause_Effect
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pgfoundry.org/pipermail/pgcluster-general/attachments/20071228/5605ca99/attachment.html