high availability and master site

[Christy Hodges]

Hello all, has anyone implemented high availability and also are using the master site feature in cloverleaf?  We are wanting to know how Cloverleaf behaves during a failover event when only one master site is set.  The type of failover I am interested in is when you have test failing over to prod or vice versa.  The master site defined on the two environments are different.  Please let me know if you have this set up and also if both environments worked properly during this event.  thanks!

Topic

[James Cobane]

From a fail-over perspective I would think that the master site defined in your prod instance would remain as such (it’s just another site) and the fact that you are failing over to your test hardware is independent of the test sites.  Your test and prod sites should exist in separate logical partitions from a fail-over perspective so it shouldn’t matter if your test partition has site1 as the master, and prod has site2.

With that said, we don’t have a master site configured yet, but prod & test LPARs are independent and shouldn’t conflict during fail-over.  All the sites in our test environment are named differently than prod (i.e. site1prod vs. site1test ), but doesn’t really matter if they were named the same/different as they are defined on separate LPARs.

Jim Cobane

Henry Ford Health

[Christy Hodges]

my understanding of HA is very limited, so forgive me, but our paths on our test and production boxes are identical (with the site names as the only difference).  during a failover, if our data is the only thing that is being failed over and master site is one per root, then i’m not understanding how there could be two master sites respected during a failover.  please unconfuse me 🙂

[James Cobane]

While the software paths are the same on test and prod (which they normally are), the test and prod instances exist within their own logical partition (LPAR).

For example, you would have 2 physical machines, each with a prod and test LPAR.  On machine 1, the PROD LPAR is normally active, with the TEST LPAR not doing anything, and conversely on machine 2 you have the TEST LPAR normally running and the PROD LPAR just sitting there.

In the event of prod failing over, the PROD LPAR on machine 2 would get mounted and become active with the prod stuff, and the TEST LPAR would continue to run, but likely at a reduced priority.

So, it’s like they are on separate machines even though they reside on the same physical piece of hardware when the fail-over happens.

I would recommend talking with your Sys Admin to understand how fail-over is configured in your environment.

Hope that helps.

Jim Cobane

Henry Ford Health

[Peter Heggie]

We could not implement a master site with HACMP; at least, it worked fine until we failed over. It was several years ago and my memory is fuzzy; I’ll try to dig up some old emails.

Our configuration is AIX, with a prod server and a test server. The sites are soft-linked from /integrator to the /prod/cis6.0/ folders. We have a /prod file system on our prod server, and a /test file system on our test server. When production fails over, the /prod file system is moved to the test server.

The /test file system is there also and both prod and test sites are running at the same time. If memory serves me, this is where the problem occurred. We had implemented a master site in our test server first. When prod failed over to test, production processing used the test master site.

That is my recollection; I will have to double check.

Peter Heggie

[Author]

Replies