Vcs cluster start issue

[symsonu]

if we have two node cluster  with nodes A and B. running service groups A on node A and B on node B

In generally speaking, starting the 2nd node B when node A is runinng with VCS and ServiceGroup,  will not cause any issue.

lets say if we we have bring the A node by force seeding with /sbin/gabconfig -x and is running SG A and B.

Now, we need to bring the second node up and form cluster.

How  should we go about  in a most proper manner with minimal outage.

[mikebounds · Accepted solution]

If node B is not available so you bring up A node by force seeding with /sbin/gabconfig -x and then start SG A and B on node A, then node B will just join normally when it comes up, see that node A has all service groups and so there is nothing you need to do.

It is WHEN you force seed a cluster you need to do something, not AFTER.  When you force seed cluster on node A, you are in effect saying, the reason node B cannot be seen is NOT because heartbeats are not working, but because node is down - i.e when gab starts on node A and it can't see B, gab on node A will not seed the cluster as it can't be sure B is not up and actually has the service groups already online and has its heartbeats disconnected, so when you seed cluster on node A, you MUST make sure node B is not up.

So the gab seeding is protecting against pre-existing split-brain, meaning the heartbeats are disconnected when the nodes start.  If when B comes up (after you have manually seeded cluster on node A), and the heartbeats ARE disconnected, then B will not seed as it can't see A and in this case you must NOT seed node B otherwise you will get split-brain and you would then have 2 nodes running their own cluster without being able to see each other.

Mike