[Pgpool-general] Setting up connections config for load-balancing

[Tatsuo Ishii]

> > Any suggestions on how I can troubleshoot this?  Very little diagnostic
> > data from pgpool appears to be available.
> 
> More on this: PCP thinks that node 1 is down:
> 
> pcp_node_info 10 localhost 9898 postgres postgres 1
> db3 5432 3 1431655764.666667
> 
> I don't know how it got this idea.  There isn't anything in the
> configuration to suggest it, and connecting to node1 via psql from the
> pgpool host works fine.

Sure. pgpool disconnects "unhealthy" DB node (in this case node 1) and
continues to work on.

> Do I have to run pcp_attach_node every time I
> restart pgpool?

No. I suspect bogus pgpool_status file is remained. Try:

Shudtown pgpool
Remove pgpool_status
Start pgpool

I recommend this time starting pgpool with -d (debug) option and see
everything is fine.

> Also, those weighting numbers are *very* odd, though.  What's up with
> that?  Is that maybe the cause of no queries going to db3?

This is normal. The value you see is the normalized value. From the
comment in pool_config.l:

	/*
	 * Normalize load balacing weights. What we are doing here is,
	 * assign 0 to RAND_MAX to each backend's weight according to the
	 * value weightN.  For example, if two backends are assigned 1.0,
	 * then each backend will get RAND_MAX/2 normalized weight.
	 */
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp