Need Info: Sun Cluster 2.2

[schu]

I would like to implement High Availability on our Sun servers but would like a bit more information first.

I have two servers, one production and one Testing. I would like both machines connected as a cluster with the production being the primary.

Sun is helpful but they only provide all the nice features, which I already know. What I want to know is...

1. When the primary node fails and the secondary node takes over, does the secondary node need to reboot, to take over IP addresses, etc.

2. If I have two nodes, is one of the nodes redundant? In other words when the primary node fails does the second node take over all the primary nodes task. If so what happens to the tasks on the secondary node.

3. Sun cluster guards against one single point of failure. If I have two primary and one redundant. is that possible. So the redundant takes over the tasks of whichever primary fails.


Thx.

 

[Guest_imported]

1. When the primary node switches over and the secondary node takes over, the process will be: primary node will unmount all the filesystems, deport the diskgroups and down the virtual ip address. The secondary will then import the diskgroups, mount the filesystems up and will plumb(create), configure and up the virtual ip address of the logical host. There is a 12 step process that it runs through but I am not sure what step corresponds with what it does. There is no need to reboot since the ip address is just downed on the primary and up on the secondary. If there is a failure on the primary server it may initiate the switchover but it would depend on the problems.
2. If I have two nodes, is one of the nodes redundant? I would assume that you have stuff running on server b that can failover to server a and vice versus so both servers are redundant. To make it truly HA is another thing. As far as what happens to the processes running on the server that takes over for the other server, it would probably run a little slower but still should be able to handle the load.
3. Sun cluster 2.2 allows up to a 4 node cluster and the failure I believe is based on the least busy server at the time gets the load. Based on the fact that you can have anyone node fail over to any of the others all 4 nodes would then be redundant. There is more to this but my cluster memory is a little fuzzy.

 

[schu]

Thanks, I think that handles most of my question, but it does lead to another question.
You mentioned deport and import of disk groups, therefore I assume there is a disk array. But how about the local disks? Do I have to maintain the same local scripts all across the other servers? That would mean there should be a copy of the local disk across all nodes? It that correct?

regards...

 

[Guest_imported]

There should be a copy of the information you need to startup your application across the nodes. The internals don't have to be exactly the same but for sanity sake it helps to keep them close.