06-14-2009 12:59 AM
I have a req where i need to config 2 node cluster on HP UX V2 . 2 servers are Cx 2600 and the storage is EVA.The req is one application runs on Node 1 and the other on node 2. If any failure on node1 then Node 2 will take over the applivation 1 and if any failover on Node 2 then Node 1 will take over application 2. my concern is
1. Will both application can write to same storage at same time ( I know only one node can write )
2. Is it reqd to have diffnt VGs?
3.Whats the criticality involved in this case.
4. Any whitepapers or any docs or any pointers please
Attached is the brief req
Solved! Go to Solution.
06-14-2009 01:09 AM
- Active/Active can mean where 2 nodes in a cluster are both running applications, but seperate applications writing to seperate storage. Microsoft would call MSCS active/active in that very often you run one SQL server instance on one node and a seperate one on the other.
- Active/Active can also mean where you have one application with processes active on both nodes, writing concurrently to the same piece of storage. Oracle RAC is the classic example of this.
In your case its sounds like you're talking about the first definition. So to answer your questions:
1. Depends what you mean by "storage" - if you mean to the same disk array then yes, if you mean to the same filesystem, logical volume, volume group or LUN then *no*. Don't even strat thinking that one node can read from a filesystem whilst the other is writing - unless you have a proper cluster filesystem you shouldn't do this.
2. Yes its required to have different VGs - and that implies different filesystems, different IP addresses and different start/stop scripts to.
3. This question makes no sense
4. Beyond what I've written above? No not really, just follow the manual for Serviceguard:
06-14-2009 01:47 AM
Thanks for the swift reply.
YEs ur right my req matches first case Where application 1 runs on node 1 and application 2 runs on node 2 And I am gng to create different packages for both the nodes. My concern was about Volume Groups Your answer cleared my doubts I need to have diffnt VGs for Diffnt packages and for diffnt nodes.
So when Appl 1 failes on Node 1 it will switch to Node 2 and When Appl 2 fails on node 2 it will switch to Node 1. But package P1 on node 1 shd write to Vg1 and Package P2 on node2 shd write to Vg2.
I will create vg 1 and vg 2 in the same EVA.
Correct me if I am wrong. Thanks once again
06-14-2009 05:59 AM
The methodology you describe will work.
Normally a package needs to activate a volume group at start up, so it will need its own volume group on shared storage
Some applications like Oracle RAC can use service guard and both access the same data at the same time from two different nodes. That requires special volume group work.
The situation you describe, packages on both nodes is simple and just requires you get the shared storage configured so each package has its down volume group to activate.
Owner of ISN Corporation
06-14-2009 07:53 PM
Whether or not both vgs are on the same storage, this will allow you to have one application running on each node at one time. At times as needed, fail over to the other node will be possible as well.
06-16-2009 09:35 PM
Thanks a lot for the inputs. I created design document with all the inputs.
I created cluster also. Hope it will work as expected.
Thanks once again