Обсуждение: Downtime

Поиск
Список
Период
Сортировка

Downtime

От
Rafael Domiciano
Дата:
Hello for all,

Our bussiness do not tolerate anymore downtime (the downtime is estimate at 10 hour / year - that involves upgrade of hardware, crash of hardware), and using Slony-I 1.2.15 I have waste time to put the slave node on production.
So I'm studying some replications softwares, like pgpool, pg_cluster, pg_replicator.

Pgpool has a good idea, I can connect at 2 servers or more at the same time. But I have some limitations, like Replication, Datas generated by the servers (CURRENT_DATE, CURRENT_TIME, serial, oid); so I discarded it.
I tryed pg_cluster, but the documentation is very poor, and the installation was unsuccesfull.

I would like to know from you what solution you uses for avoiding downtime, so I could choose the solution that is best to my scenario and for others that is begining in Postgres and is a little lost in replication.

Thnks for all,

Rafael Domiciano
DBA Postgres

Re: Downtime

От
Scott Marlowe
Дата:
On Mon, Feb 16, 2009 at 12:52 PM, Rafael Domiciano
<rafael.domiciano@gmail.com> wrote:
> Hello for all,
>
> Our bussiness do not tolerate anymore downtime (the downtime is estimate at
> 10 hour / year - that involves upgrade of hardware, crash of hardware), and
> using Slony-I 1.2.15 I have waste time to put the slave node on production.
> So I'm studying some replications softwares, like pgpool, pg_cluster,
> pg_replicator.

What exactly was not good enough about slony?  We use it for automated
failover which is initiated by the application if it detects the
master go down.  Works pretty well.  Our downtime for unscheduled
maintenance is well under an hour a year with this setup.

Re: Downtime

От
Rafael Domiciano
Дата:
Hi, Scott

Slony is a good software at all, we are using it now to replicate real data to a dedicate report server and to a data center and works very well.
But when we have some downtime like crash hardware, the time to put the slave node running, even as of less of 1 minute is too long for the company bussiness, loosing a great amount of sales authorizations.
So we need a solution that in the time of crash another server right in the moment could be working, receiving connections and transactioning.

This week I installed and used pg_cluster. I liked the idea of the software, and is that what i'm looking for. But pg_cluster seems that is no more upgraded, and there is a lot of bugs and errors.
I'm trying now the CyberCluster from CyberTech, developed under pgcluster.

I hope that you could understand the scenario.

Thnks,

Rafael Domiciano

On Mon, Feb 16, 2009 at 5:01 PM, Scott Marlowe <scott.marlowe@gmail.com> wrote:
On Mon, Feb 16, 2009 at 12:52 PM, Rafael Domiciano
<rafael.domiciano@gmail.com> wrote:
> Hello for all,
>
> Our bussiness do not tolerate anymore downtime (the downtime is estimate at
> 10 hour / year - that involves upgrade of hardware, crash of hardware), and
> using Slony-I 1.2.15 I have waste time to put the slave node on production.
> So I'm studying some replications softwares, like pgpool, pg_cluster,
> pg_replicator.

What exactly was not good enough about slony?  We use it for automated
failover which is initiated by the application if it detects the
master go down.  Works pretty well.  Our downtime for unscheduled
maintenance is well under an hour a year with this setup.