Hi Detlef et al,
can you expand a bit on the resilver part?
When/why would resilvering be necessary, and when would that be a manual
process and not part of the automated behaviour of the agent (since you
mention example scripts)?
Greets
Thorsten
Detlef Ulherr wrote:
> Hello everyone,
>
> I would like to propose to enhance the functionality of the PostgreSQL
> agent.
>
> Terminology:
>
> I am using the term cluster under two meanings:
>
> 1.
>
> OHAC Cluster which is the high availability software, here the
> term cluster is alwaays preceded by OHAC.
>
> 2.
>
> PostgreSQL cluster which is a set of postgres databases tied
> together by a set of postmaster processes, in this usage the term
> cluster is always preceded by PostgreSQL or database.
>
> Currently PostgreSQL 8.2 allows data to be replicated between two
> PostgreSQL databases clusters by shipping it's write ahead logs (WAL
> logs) from a primary active database to a remote standby database cluster.
>
> This implies that only one PostgreSQL database cluster is active and the
> remote database cluster is running in recovery mode and applying the WAL
> logs that were shipped from the primary database.
>
>
> With the current version of the PostgreSQL agent it is only possible to
> use this functionality to ship the logs from a PostgreSQL database
> cluster under OHAC control to a standalone system not under OHAC control.
>
>
> My proposal would be to enhance the agent with two options:
>
> 1.
>
> The PostgreSQL agent should be able to control a database which is
> running in recovery mode. In recovery mode, the monitoring has to
> work on a reduced level. If the recovery mode terminates and the
> database transfers into ready state, the agent should switch to
> full monitoring without intervention.
>
> This will allow the WAL log shipping functionality between two
> different clusters.
>
> 2.
>
> The agent should allow the the log shipping between two databases
> within one OHAC cluster in a shared nothing topology. If the
> primary database fails, the standby should be converted to a
> primary without intervention.
>
> This will allow PostgreSQL to be used under OHAC control if shared
> storage is either too expensive, or in a OHAC metro cluster where
> the latency of host based mirroring has a sever impact on database
> performance.
>
> To make the solution more comprehensive, the modified agent should
> contain example scripts which make it easier to resilver an old primary
> database after a failover while using WAL log shipping.
>
> Combining the log shipping functionality of PostgreSQL and OHAC cluster
> will give us the option to use the rich set of dependencies which may
> span multiple nodes and zones together with a replicating database.
>
>
> Many thanks in advance for considering this proposal.
>
> Regards
> Detlef
