Greetings!
In an attempt to satisfy our client's requirements for redundancy, we
ended up with an unusual (at least to my inexperienced eyes) system
configuration. Our client has two buildings that use our application.
Each building has an application server and a storage server. All are
runnning Windows Server 2003. An iSCSI controller running on a
building's storage server is configured to map an image on the D drive
of that server to appear as the H drive on the building's application
server. So, anything written to the H drive by the application server
actually is stored in STORAGE\\d:/DataStorage/data.img. PostgreSQL 8.4
is running on the application server, configured to use the H drive to
store data. Therefore, data is actually being stored on the storage
server's D drive.
The storage servers for each building also have the task of backing up
data for the other building. Symantec's Backup Exec is used for that
purpose. But we found that when the backup job that was supposed to
back up one building's storage server ran, the PostgreSQL service on
that building's application server stopped. We tried excluding the
folder containing the image used as the H drive from the backup, but it
didn't help.
Has anyone else seen a Backup Exec backup job stop a Postgres service?
Can anyone suggest ways to avoid the problem? (I'm already fighting to
run PostgreSQL on the storage server instead of the application server.
I see no benefit to having the service running on one machine and
storing data on the other. If a terrorist blows up the storage server,
it will take all of ten minutes to re-install PostgreSQL.)
RobR