[Pgcluster-general] PGReplicate: fork failed & trouble with load balancer and db_3

[Antony Petropavlovsky]

Hello.

I setup PGCluster 1.5.0rc16 for next configuration (on FreeBSD 6):

    serv_1                  serv_2                 serv_3
 -------------          -------------          -------------
|    lb       |        |             |        |             |
|    pgrepl   |--------|    db_2     |--------|    db_3     |
|    db_1     |        |             |        |             |
|             |        |             |        |             |
 -------------          -------------          -------------

Then testing, testing, reading mail-list, reading mail-list, reading mail-list .......
And all work OK! 

But my pgreplicate some time write to DEBUG:

-----------------------------------------------------------------------
.....
2007-05-01 23:18:45 [81498] DEBUG:start thread_send_cluster()
2007-05-01 23:18:45 [81498] DEBUG:sem_unlock[1]
2007-05-02 08:04:09 [72150] ERROR:PGRcreateConn():Retry. h_errno is 0,reason is 'could not fork new process for connection: Resource temporarily unavailable
'
2007-05-02 08:04:10 [72150] ERROR:PGRcreateConn():Retry. h_errno is 0,reason is 'could not fork new process for connection: Resource temporarily unavailable
'
2007-05-02 08:04:11 [72150] ERROR:PGRcreateConn():Retry. h_errno is 0,reason is 'could not fork new process for connection: Resource temporarily unavailable
'
2007-05-02 08:04:12 [72150] ERROR:PGRcreateConn():Retry. h_errno is 0,reason is 'could not fork new process for connection: Resource temporarily unavailable
'
2007-05-02 08:04:13 [72148] ERROR:replicate_loop():fork failed (Resource temporarily unavailable)
2007-05-02 08:04:13 [79870] DEBUG:quick_exit:signo = 15
2007-05-02 08:04:13 [79778] DEBUG:quick_exit:signo = 15
2007-05-02 08:04:13 [72160] DEBUG:quick_exit:signo = 15
2007-05-02 08:04:13 [72159] DEBUG:quick_exit:signo = 15
-----------------------------------------------------------------------

The last access to any db was at 23:18. And pglb stopped manually at 23:18.
In the system log:
May  2 08:04:07 files kernel: maxproc limit exceeded by uid 70, please see tuning(7) and login.conf(5).
May  2 08:04:13 files last message repeated 6 times

In login.conf maxproc=unlimited for all.
Pls, help me with this problem.



And the next trouble I have with load balancer and db_3.
In pglb.conf: Max_Connect = 10 to any db. (In postgresql.conf: max_connections = 100)

When I start pglb it wite:

2007-05-01 23:18:36 [81488] DEBUG:PGRset_status_on_cluster_tbl():host:db_1 port:5432 max:10 use:0 status1
2007-05-01 23:18:36 [81488] DEBUG:PGRset_status_on_cluster_tbl():host:db_2 port:5432 max:10 use:0 status1
2007-05-01 23:18:36 [81488] DEBUG:PGRset_status_on_cluster_tbl():host:db_3 port:5432 max:10 use:0 status1

I run:
pgbench -h serv_1 -p 5433 -S -c 20 -t 30 pgbench
It works ok. In log of postgresql I see info about 10 connection to db_1, db_2.

But if I run:
pgbench -h serv_1 -p 5433 -S -c 25 -t 30 pgbench
I see 10 connection on db_1, db_2. But no one to db_3!
And give a msg: Sorry, backend connection is full.

In DEBUG of pglb I can`t find any msg about work with db_3.

If I change order of dbN in pglb.conf to db_1, db_3, db_2 it work with db_1 and db_3 but not with db_2.
pgreplicate work with all db.

-- 
 Anton