[pgpool-general: 2301] Re: Many short CGI connections - needed to downgrade to version 2.3

[Tatsuo Ishii]

>>> 2. Subsequent sessions to a pgpool process SELECTing from same table only caused a check to see if it was temporary, by way of 2 commands that took 500ms before SELECT was sent to slave.  
> 
>>You could eliminate temp table checking by:
>>check_temp_table = off
> 
> This option has reduced subsequent selects in a pgpool process from 500ms to 250ms. I think it still tries to set environment variables on master before the first query of each client connection. 

Yes, pgpool needs to know if the query contains writable functions.

>>If a query is sent right after the table is created, querying to slave
>>may fail because the creation of the table might not be propagated to
>>slave yet.
>>
>>Maybe we can eliminate all of the checking if user agree on the risks
>>(that's what pgpool-II 2.3 behaves). Do we want to add such a switch?
> 
> Either this, or can we stop it from setting variables on the master(user=xxx,db=xxx,application_name=xxx), unless the query is to be passed to the master? That would solve the problem without need for more options. 

At this point, no. We might implement it as a new feature for
pgpool-II 3.4 if there's enough interest.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp