[Pgpool-general] pgpool locking

[Tatsuo Ishii]

> I'm using pgpool 2.2.2. and wanted to know if the following is expected
> behavior.
> 
>  
> 
> Configuration:
> 
> +---------------------------------+
> 
> | (Server A)                      |
> 
> | App ---> pgpool -----> backend0 | 
> 
> +----------------\----------------+
> 
>                   \
> 
> +------------------\--------------+
> 
> | (Server B)        \             |
> 
> |                    `-> backend1 |
> 
> +---------------------------------+
> 
>  
> 
> Test:
> 
> 1. Only Server A is running. Server B is down.
> 
> 2. Start pgpool.
> 
> 3. Try connecting to pgpool (e.g. pgsql -p 9999). At this point, client
> (psql) holds. I think it's at TCP level. However, I changed the tcp
> timeout values to be few minutes/seconds.
> 
> 4. After the timeout, pgpool is then unlocked and allow connections.
> 
>  
> 
> I can see pgpool doing this by design. I just wanted to know if it's
> this way and if would be possible to have pgpool just degenerate any
> backend that is not available within some configurable period of time.

Have you enabled health checking? It should be able to detect this
network failure.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp