cancel
Showing results for 
Search instead for 
Did you mean: 

cvm

tanislavm
Level 6
hi, cvm offer support for raw devices(raw logical volume) and for cfs.rigt?if we have raw logical volume only,then we need to install cfs foundation as in the case that we use only cfs? raw logical volume option is better than cfs when there is a very large I/O.right?So when is better to chose cfs over raw logical volume?beside the case that the application not support raw logical volume. let`s say we have 2 nodes vcs with cfs on it(so parallel cluster),one instance(of the same DB) runs on node a and one instance(of the same DB) on node b.The 2 nodes share the same dg. If node a has problem with path disks and is not able to access anymore the whole dg,what happens next? What happens if the dg come online next? In HA failover vcs is there a way to automatically fail back a service group,who failed over on other node? thank so much, marius
1 ACCEPTED SOLUTION

Accepted Solutions

Gaurav_S
Moderator
Moderator
   VIP    Certified

Hi Marius,

Yes CVM provides support for raw volumes. CVM is embedded in volume manager (vxconfigfd) binary, if you enable CVM/CFS license you can use them. While you are running the installer script, you will need to choose the right option to install CVM/CFS

In the scenario you have mentioned, what will happen if node A has problems, this would depend on what role node A had. If node A was master then your DG would disable & volume should report IO errors, this will effect the DB instances running & corresponding service groups should fault

If your node A was slave (node B was master) if your node a has problem with disk paths, parallel DB instance running on node A should fail while your node B should still continue to access as path from node B are alive & active & DG is accessible

For your next question, you can tune multile parameters (systemlist, autostartlist) however why would you want to have an autofailback ? Imagine a setup of 2 node, node A & B. Lets say node A faults & your service group moves to node B. Now your node A comes back online, users using the application in service group will face an outage if failback happens . so rather leave it for a proper outage window & do manual failover.

 

G

View solution in original post

2 REPLIES 2

Gaurav_S
Moderator
Moderator
   VIP    Certified

Hi Marius,

Yes CVM provides support for raw volumes. CVM is embedded in volume manager (vxconfigfd) binary, if you enable CVM/CFS license you can use them. While you are running the installer script, you will need to choose the right option to install CVM/CFS

In the scenario you have mentioned, what will happen if node A has problems, this would depend on what role node A had. If node A was master then your DG would disable & volume should report IO errors, this will effect the DB instances running & corresponding service groups should fault

If your node A was slave (node B was master) if your node a has problem with disk paths, parallel DB instance running on node A should fail while your node B should still continue to access as path from node B are alive & active & DG is accessible

For your next question, you can tune multile parameters (systemlist, autostartlist) however why would you want to have an autofailback ? Imagine a setup of 2 node, node A & B. Lets say node A faults & your service group moves to node B. Now your node A comes back online, users using the application in service group will face an outage if failback happens . so rather leave it for a proper outage window & do manual failover.

 

G

tanislavm
Level 6

Hi,

Please also give me your comment on my new discussion "vcs".