Hi everyone,
We're currently testing pgcluster and we're having trouble with apparent
deadlocks when running concurrent inserts. I've looked through this
mailing list and have seen posts from others that have experienced the
same problem.
Based on these posts, I've tried setting auto_lock_table to false. It
seemed to help some, but didn't completely eliminate the problem. And
this is with only five test threads; we were hoping to deploy this to a
production environment which could have 100 or more concurrent database
connections.
I guess I have two questions for everyone:
First, are others still having problems with concurrent inserts/updates?
Or is it that I'm just doing something wrong?
Second, is anyone using pgcluster in a high-traffic, mission-critical
situation? Or is it maybe not quite ready for that yet? We'd like to
get a feel for how stable it is long term, and knowing in general how
many people are using it would help.
Here's a little more info on my test setup:
We're using pgcluster-1.7.0rc12. 2 cluster servers, 1 replication
server, and one loadbalancer. (The replication server and the
loadbalancer are actually running on the same machine; the 2 cluster
servers are on separate machines.) The OS is CentOS 5.1 (32 bit x86) on
all machines.
If I run just one test thread then everything seems to work OK. Also,
if I restrict the loadbalancer to only know about 1 cluster server then
running multiple tests works. However, if I configure the loadbalancer
to see both cluster servers and run with multiple test threads then I
get deadlock.
When the deadlock happens, I can look at the currently executing queries
on both cluster servers and see inserts in progress on the same table.
Thank you for your help,
Michael
_______________________________________________
Pgcluster-general mailing list
[email protected]
http://pgfoundry.org/mailman/listinfo/pgcluster-general
