02-11-2014 05:48 AM
Hi,
I have looked into my work VCS environment. Some VCS cluster nodes got VIP setup so I know when one of the node failed or exit, the virtual IP will failover to another node so that the client connection will still be connected to the service. However, I do see some two node clusters which do not have the VIP setup and the service group are both showing online each node. How does this failover happends in this situation?
Thanks
02-11-2014 05:56 AM
If you have a service group which is showing online on both nodes, then this is a parallel group and it does not failover, so VCS is just controlling or maybe just monitoring the resources to ensure they are online and can be optionally configured to restart the resources in this group, should they fail.
It is quite common to put network resources in a parallel group, so here VCS just monitors the network cards.
Mike
02-11-2014 06:33 AM
Hi Mike,
I am not quite follow... do you mind if you can elaborate more on the purpose of setting up paralle service group on hardware resource?
02-11-2014 06:41 AM
An application needs a network card in order to function, but unlike, say a virtual IP (VIP) which can only be online on one node, so VIP fails between the nodes in the clusters, a network card is present on all nodes, so you do not failover a network card, but VCS should still monitor it, as if for instance you have 3 nodes in a cluster and node A where the application is running fails and the NIC on node B has failed and the NIC is ok on node C, then VCS will choose node C if the application is configured to require a NIC (which most do)
Mike
02-11-2014 06:28 PM
Hi Mike,
By looking at the hastatus -sum, the online services on both nodes are cvm. So the application needs the Cluster Volume Manager on either node in this case?
-- SYSTEM STATE
-- System State Frozen
A host01 RUNNING 0
A host02 RUNNING 0
-- GROUP STATE
-- Group System Probed AutoDisabled State
B cvm host01 Y N ONLINE
B cvm host02 Y N ONLINE
02-11-2014 10:30 PM
Hi,
CVM enables parallel clustering for applications which can run parallely on multiple nodes, for e.g Oracle RAC.
In above cluster you would have some application/ database running which is capable to run on multiple nodes & hence you will need cvm group to be online on both the nodes at the same time in order to use application in parallel mode.
cvm group usually consists of volume manager daemon (cvm_vxconfigd), fsck daemon (cvm_vxfsckd). These are mandatory to be running for cvm to work.
G