RAID-0 Volume Question--HP4300

[419Bob]

I have an HP 4300 cluster (2-node).  I have several raid-10 luns and 2 raid-0 luns set up on this cluster.  Each node has its nic's bonded and set up for ALB.  When I set this cluster up, I didn't realize that Flow Control is turned on by default.  I'm experiencing some slowdowns with this particular cluster and I believe flow control needs to be shut off.  (My other 4300 cluster runs fine and...so far as I can tell, this is the only difference between the 2 clusters.)

Can I simply remove the servers that are accessing the RAID-0 luns and then break the nic bonds and make the changes without losing any data on that LUN?

I'm also assuming that any servers accessing the RAID-10 volumes will continue to be able to access them during this change?  Is this correct or do I also need to shut off all access to this 4300 cluster?

Thanks!

[Bryan McMullan]

That's correct, just disconnect the servers accessing the RAID-0 volumes and perform your changes.  The RAID-10 volumes should be fine, though you may want to ensure that the paths reconnect after updating one of the nodes before you start the second one.  That should happen automatically, but I've seen it take a few minutes in the past. 

If that doesn't fix the issue,  you may want to contact support.  The issue may be drives that are failing, but just haven't hit the point where they are considered "bad".  A number of times I had issues with a cluster and it turned out that multiple drives were failing at one time.  This was usually indicated by the node throwing "overloaded" errors.

Good luck!

[oikjn]

I have yet to see a document suggest turning OFF flow control.  Better to keep it on and make sure the rest of your infrastructure has it enabled.  You can have slowdowns when its only partially on some devices and not others, but the solution is not to turn it off, but turn it on for everything.

[Jess1313]

Having Flow Control turned on in itsels SHOULDN'T cause slow downs, with that said it could be making it visible to you that you have some poor performing componenets of your infrastructure. 

With that out of the way, yes your assumptions of disabling access to the NR-0 volumes, and leaving the NR-10 up are fine. 

If your using a FOM ensure its not stored on SAN storage as you will learn when you power down how important that is, and if your not using a FOM make sure your actions will not cause a quorum loss (CMC > Cluster > Storage node you will power down > availability tab)

[silus]

I think we may also have to disable flow control on the nics for performance as we don't have it enabled on the switches/esx hosts.

If we break the alb bond, will it automatically assign the bonds IP to a NIC or will it drop the IP entirely and we will  have to manually set it at the console?

[Bryan McMullan]

One of the NIC's should take the IP of the bond, the other is typically set to 0.0.0.0.  It sometimes takes a few searches for the system to be responsive after breaking the bond, but it should respond within 30 seconds or so.