[Pgpool-general] Clarifying interaction between master/slave mode and automated failover [SumsaultRT #3219]

[Tatsuo Ishii]

> Hello,
> 
> Thanks for your work on pgpool. I'm currently researching using it in conjunction with Slony-I
> for load-balancing.
> 
> I'd like to clarify some points from the documentation:
> 
>  1. No automated failover will happen unless "health_check_period" is set. Correct?

No. If fatal errors such as communication problem with a server
occurs, automated failover will happen.

>     Another parts of the documentation indication that failover is
> 	automatic rather than that it /can be/ automatic.
> 
>  2. What's the behavior when failing over from master to slave? Does it 
>     still continue to send only SELECTs to the slave, or does it began
> 	to treat it as a master? 

Basically it stops to operate in master/slave mode. Thus it will sned
non SELECT quries to the slave.

> I would like for pgpool not to start treating the slave the master,
> because I'm afraid that could mess things up if there is a network
> blimp, but the Slony-I master/slave replication is working OK.  
> 
> This part of the switchover documentation could be clearer:
> 
>    $ pgpool [-f config_file] [-s {m[aster]|s[econdary]] switch
> 
>    If [-m {s[mart]|f[ast]|i[mmediate]}] part is omitted, master goes
>    down. To stop secondary try:
> 
>    $ pgpool [-f config_file] -s s[econdary] switch
> 
> It says "If [-m {s[mart]|f[ast]|i[mmediate]}] ...", but those options
> aren't included in the example above it. 
> 
> My guess is that:
> 
>  # turn off the master
>  $ pgpool -s master switch
> 
>  # turn off the secondary (possibly a slave)
>  $ pgpool -s secondary switch
> 
> Is that correct? 
> 
> Thanks!
> 
> 	Mark
> _______________________________________________
> Pgpool-general mailing list
> Pgpool-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgpool-general
>