P4500 cluster - re-configuring disk RAID & network RAID

[Robert Neild]

We have three P4500 G1 5.4 TB SAN servers in a high availability cluster, in a single rack.  The SAN cluster grew from a single unit with disk RAID 0+1, to a pair using network RAID 0, to three configured with redundancy with network RAID 5.  Because of how the cluster evolved, we are now configured with disk RAID 0+1 and all LUNs at network RAID 5. 

Due to performance reasons and recommendations from HP tier 3 support, we now would like to re-configure the cluster so we are on disk RAID 6 and all LUNs at network RAID 0+1.  From what I've read, this is going to be painful to achieve.  All LUNs are thin provisioned.  According to the CMC, we have total space of 7,409 GB and approximately 1,830 space available.  The clients are a mix of VMware VMFS and Windows server-mounted volumes. 

To simplify things, I've already upgraded all of the P4500s to the latests firmware and SAN/IQ 9.5 over the past week. 

The documentation indicates that I have to remove the P4500 servers one at a time from the management group, re-stripe from disk RAID 0+1 to RAID 6, then re-add the server to the management group and cluster.  Before I can do this, it appears that I must re-stripe my LUNs to fit on two of the P4500 servers while the third server is being re-striped.  It looks like converting my LUNs to network RAID 0 is my only alternative considering the amount of available space on the SAN cluster. 

My concern is whether the network RAID 0 LUNs will safely be retained intact when I remove a P4500 from the management group.  Has anyone gone through this process before that can give some advice?  I thought I'd try here prior to opening up an incident with HP support. 

Thanks in advance! 

[Bryan McMullan]

I've not gone through exactly what you are going through, but I have changed the disk raid on a cluster before.  You're steps seem to be correct for the most part. 

If you look at one of the nodes in the CMC and click the Availability tab, does it show that you'll lose any volumes if one node goes down?  If so, you're in quite the pickle and the process is going to take you quite a bit of time. 

1. Restripe to RAID 0 for non-critical volumes until available space is below 2 units.

2. Remove 1 Node from management group (wait for eons while data is migrated off)

3. Apply RAID 6 to disk.

4. Add Node Back to cluster (wait for eons while data is migrated back)

5. Repeat steps 2-4 for remaining 2 nodes.

As long as you don't have any outages, the above should work fine.  Be prepared for this to take quite a while.  At least multiple days, most likely longer.  When I switched from RAID 10 to RAID 5 on a cluster it took over a week. 

[ccavanna]

Don't forget to crank the Bandwidth prioirity on the management group.

[David_Tocker]

Another possible solution would be to break out the VSA trials to perform the migration - if you have some spare servers with storage sitting around this may prevent some panic on the migration... Keep in mind that if it is heavily loaded you may loose some performance during this process depending on the hardware.

[AUSSUPPORT]

Hi ,

  I got 2 x P4500 and 1 x VSA in DR site. Each node got 7 TB RAW after doing RAID 5 on disk and network RAID 10  i got 5 TB usable space?

 So what is the best way to setup this to make more space?

AS