[Pgcluster-general] 1.9.0rc5 deadlocks

[Stan Carney]

Hello All,

I'm in the process of moving a Java (Hibernate + JDBC) application I 
created over to PGCluster. It runs well on regular postgres but once I 
move to PGCluster I'm unable to perform inserts within a transaction. It 
appears to lock at the exact same place everytime. select * from 
pg_stat_activity shows '<IDLE> in transaction'. This problem doesn't 
happen with regular postgres.

My setup is as follows but it really doesn't appear to matter as I can 
shutdown everything but the bare minimum to run (i.e. db0) and I still 
get the same issues. pgcbench (pgcbench -i dev; pgcbench -c 90 -t 1000 
dev;) works as expected in all cases.

app0: load balancer (lb0)
app1: load balancer (lb1)
db0: replicator (rep0), cluster 0
db1: replicator (rep1), cluster 1

I can post the entire configuration if required but I'm pretty sure the 
entire stack of applications is working as pgcbench works as expected.

The insert statement the application hangs on is a prepared statement 
into a table that has a composite primary key. I'm able to reproduce 
this 100% of the time. If I comment out the code that calls the prepared 
insert statement the exact problem is encountered later on the same table.

Thoughts?


Thanks,
Stan