[Pgpool-general] Basics of replication with pgpool-II - how does it work? compared to Slony? to Postgres log shipping?

[Aleksey Tsalolikhin]

Dear Tatsuo,

  Thanks, that's a great explanation, very clear.

  Maybe you could add it to the pgpool-II manual, in the
Replication section?  Just a suggestion.

Thank you!
Aleksey


On Mon, Jun 7, 2010 at 6:00 PM, Tatsuo Ishii <ishii at sraoss.co.jp> wrote:
>> Hi.   We want to have redundancy for our database;
>> is pgpool-II the right tool, with it's replication mode?
>>
>> We want to have 2 database servers, and if one
>> dies, our application just goes on using the other
>> one, while we repair the broken one.
>>
>> Can pgpool-II replicate writes such that if a write
>> query is submitted, pgpool-II will write it to DB server
>> A and DB server B, and will only return a success
>> to the client if both writes succeeded?  otherwise
>> it will return a failure, perform any necessary rollback
>> if the write was replicated only to one server?
>
> Yes, pgpool-II exactly works as you described. Plus, with write query
> pgpool-II start a transaction if it's not already started by user.
> The write order is first A then B (if A is the first one on
> pgpool.conf). Commit order is B then A.
>
>  If write query fails, pgpool-II aborts the transaction and start
> a  degenration, which will be seen as connection shutdown from
> PostgreSQL server by client. By reconnecting pgpool-II, client can use
> the healthy PostgreSQL.
>
> Please note that "fail" does not mean here a query error (for example,
> violation of constraints). Rather "fail" means network error or
> PostgreSQL server down or some such.
>
>> If that's not how pgpool-II replication works, how does
>> it work?  I've looked at the manual but I still don't
>> understand.
>>
>> I'm familiar with replication with Postgres 8 log shipping
>> (warm standby),  Postgres 9 streaming replication
>> (hot standby), and Slony (hot standby).  How does
>> Pgpool-II replication work and how does it compare to
>> the Postgres/Slony replication?
>
> They are all log(physical or logical one) based asynchronous
> replication systems. Pgpool-II is a query based synchronous
> replication.  That's the big difference. In addition, those
> replication systems do not have any automatic failover/load balancing
> functionality while pgpool-II has.
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese: http://www.sraoss.co.jp
>