Обсуждение: Clustering of pgsql 9.2

Let me begin by deferring to The Big Brains, but I am not sure there's enough in this to really advise you from The Implementation Guys. That's me. I/O for any database is an issue, and when I hear any db talking to a SAN, I get a bit leery. Couple that with replication, HA and failover?

If you could explain a bit more about what you're trying to accomplish, what expected performance should be, how that SAN is actually utilized, and such, I might be able to provide some experience on the issue.

Hi,

On Thu, 2015-01-15 at 05:24 +0000, shyamkant.dhamke@wipro.com wrote:

> We have following environment for  database clustering at production environment:
>
> 1.       Postgres is running on RedHat 6 enterprise edition for primary and failover server
>
> 2.       SAN disk is available
>
> Now I need help on following
>
> 1.       What should be clustering software used for Redhat clustering ? Can Veritas Clustering Software (VCS)
clusteringbe used for clustering REDHAT OS? 
>
> 2.       How the SAN disk can be used to store the data centrally? Can this be used as QUORAM disk?
>
> 3.       Is there any service for Postgres Database service clustering?

http://www.gunduz.org/download.php?dlid=208

This presentation gives you an overview about how to setup Red Hat High
Availability Addon (formerly Red Hat Cluster Suite) with PostgreSQL,
which matches your needs.

Regards,
--
Devrim GÜNDÜZ
Principal Systems Engineer @ EnterpriseDB: http://www.enterprisedb.com
PostgreSQL Danışmanı/Consultant, Red Hat Certified Engineer
Twitter: @DevrimGunduz , @DevrimGunduzTR

Looks like a large size database. However, I would be concerned about the throughput on a SAN. Also, you didn't specify the replication/failover stuff.

So...never replicate or fail over to the same hardware. I disagree with the concept of databases on SANs (and we can open THAT discussion in another thread). I will pretty well tell you that no one here has my IT/HA/OS experience. The Big Brains sure do know a whole lot more about PG, that's certain.

However, you are talking HA and failover (those are separate) and replication  (also separate).

I would love to advise you on this list where the Big Brains are watching.

So...

What's your primary server?
Why does it need to replicate?
Is PITR an option?
Is PITR a necessity?
How many servers are involved?
What's your bandwidth between master and slave(s)
?