[pgpool-general: 9085] Re: [External] Re: Suggestions on backend_clustering_mode

[Tatsuo Ishii]

>> >> Another workaround is to configure PostgreSQL's synchronous replication to avoid replication delay.
>>
>> I'll look into this. How does the performance of this option compare with the snapshot isolation mode in pgpool?

> I did a fair amount of reading on the snapshot isolation mode. It looks like this mode sends all commands to all backends and postgresql doesn't need to do any replication.
> Am I reading that correctly?

Yes, you are.

> If this has good performance this sounds like a simplier way to do the clustering.

Unfortunately no. Due to the over head of transaction management in
pgpool, performance of the snapshot isolation mode is worse than the
streaming replication mode (sync).

> What are the reasons that people don't use this by default?

Some SQLs need special care. See the documentation for more details.
https://www.pgpool.net/docs/latest/en/html/restrictions.html

Another reason is only repeatable read transaction isolation level can
be used in the snapshot isolation mode.

Best reagards,
--
Tatsuo Ishii
SRA OSS LLC
English: http://www.sraoss.co.jp/index_en/
Japanese:http://www.sraoss.co.jp