Can the RT database be failed-over to an alternate site?

Hello,

We are about to implement RT as a core service and want to add its DB to an existing Postgresql cluster which we fail over to an alternate site when we think that the DB service will be unavailable for more than 2 hours due to power loss or network failures or some other unfortunate event. The Postgresql write ahead logs are rsynched to the alternate site and applied to the DB there which is in stand-by mode. When we promote it to the primary we change DNS service name entries to point to it. It takes about 30 minutes to execute a fail-over.

We will be running RT 4.0.6 on CentOS 6.2 x86_64 and the database will be Postgresql 8.4.7 on RH Linux.

Has anyone run RT in this kind of environment?

best regards and thanks in advance,

L. K. Carson
San Diego Supercomputer Center

Hello,

We are about to implement RT as a core service and want to add its DB to an existing Postgresql cluster which we fail over to an alternate site when we think that the DB service will be unavailable for more than 2 hours due to power loss or network failures or some other unfortunate event. The Postgresql write ahead logs are rsynched to the alternate site and applied to the DB there which is in stand-by mode. When we promote it to the primary we change DNS service name entries to point to it. It takes about 30 minutes to execute a fail-over.

We will be running RT 4.0.6 on CentOS 6.2 x86_64 and the database will be Postgresql 8.4.7 on RH Linux.

Has anyone run RT in this kind of environment?

best regards and thanks in advance,

L. K. Carson
San Diego Supercomputer Center

Hi Leonard,

We have a similar setup but not with a version of PostgreSQL quite so
ancient and we used Slony to replicate to the standby DB server when
we were using 8.4.x. We currently use PostgreSQL 9.1.x with its
streaming replication support. Way too easy to setup, run, and failover.
PostgreSQL has made tremendous strides in ease of use for its built-in
replication support. I really recommend going to 9+ which included it.
To provide a simple, minimal impact upgrade, we used Slony to replicate
from our 8.4.x instance to our 9.1.x instance. Then when they were in
sync, shutdown RT, failover, and restart RT on the new version. It took
about 30 seconds.

Let me know if you have any questions.

Regards,
Ken