|
|
Dave Hamblin
|
|
Oct 30, 2009 15:16:25 GMT
|
|
Our existing LHSan cluster is at SanIQ7. Its a 3.4 TB unit that has about 300 GB open.
Our new units (seperate cluster and managment group) are SanIQ8.1. They came in as 8.0, I upgraded to 8.1 which went fine and very fast, but they also have no volumes or data defined yet.
My question is upgrading the two 7.0 to 8.1. Can anyone tell me if this also is a fairly quick process and if there is any gotcha's or the like? I wasn't sure if the amount of data on the units plays any role in upgrade time or difficulty?
|
|
|
Note: If you are the author of this question and wish to assign points to any of the answers, please login first.For more information on assigning points ,click
here
|
|
|
Sort Answers By:
Date or Points
|
|
teledata
|
|
Oct 30, 2009 19:31:22 GMT
0 pts
|
|
|
Are you migrating the data off to your new cluster, or do you intend to do an "in place" upgrade? |
|
|
Dave Hamblin
|
|
Oct 30, 2009 19:34:00 GMT
N/A: Question Author
|
|
The new san is for new data/servers.
The existing data will remain on the existing LHSan, so it would be an inplace upgrade.
We were going to upgrade the existing units so they were current, or is this one of those "let sleeping dogs lay" (ie we should leave them alone). |
|
Mark...
|
|
Oct 31, 2009 13:51:47 GMT
3 pts
|
|
Hi,
As far as I know it is painless and can be done online. You should always read any release notes first which should advise you of any gotchas!
However, I believe when you do the upgrade it is recommended to only do one node at a time and make sure that you do the node that is serving out the VIP LAST.
Mark... |
|
Gauche
|
|
Oct 31, 2009 14:01:56 GMT
5 pts
|
|
Doing one node at a time and doing the vip last is something the upgrade process automates.
The tip to read the release notes is a good idea as their might be a patch that is suggested for you to install before the upgrade.
The size of data in the SAN does not impact how long it takes to apply the upgrade at all.
The number of nodes, and how much data you write to the SAN during the upgrade, both affect how long it takes.
Number of nodes matters because it does one node at a time.
How much you write to the SAN matters because the upgrades will wait for all volumes to re-syncronize between node reboots, and if you write 1GB of data while a node is rebooting, you'll have to wait longer for the 1GB to re-sychronize before the next node can start installing.
The less load the system is under during the upgrade the faster it will go. |
|
|
Dave Hamblin
|
|
Nov 2, 2009 12:15:50 GMT
N/A: Question Author
|
|
|
Thanks all! I will go through the release notes to make sure I am ready for the upgrade. |
|
|
Bryan McMullan
|
|
Nov 2, 2009 15:31:39 GMT
9 pts
|
|
I just upgraded our 50 nodes from 7.0 SP1 to 8.1. It went pretty smooth. The path to follow is:
1. Install Updated CMC.
2. If applicable, upgrade the DSM's for MPIO. This is a pain as it requires the old one to be uninstalled and is multiple reboots (if you are going from 7 to 8, not bad if from 8 to 8.1).
3. Upgrade all Remote Clusters (normal -> managers -> VIPs).
4. Upgrade normal nodes (non manager/non VIP).
5. Upgrade managers
6. Upgrade VIP.
NSM160's took about 18 minutes per upgrade. x3650's took about 6 minutes per upgrade. DL320/P4000's took about 6 minutes per upgrade. This is with about 30TB of data during production hours.
We did take all our connected servers offline for the manager/VIP updates as the systems wait 15 seconds I think before they hand off if you down the VIP (causing iSCSIprt errors on Windows and lockups on ESX). As we are very data intensive, I find it safer to go this route.
Good luck! |
|
|
Dave Hamblin
|
|
Nov 2, 2009 15:33:28 GMT
N/A: Question Author
|
|
|
|
Dave Hamblin
|
|
Nov 2, 2009 16:36:11 GMT
N/A: Question Author
|
|
|
Hmm, I don't see the release notes on the HP site where I downloaded the 8.1 upgrade; anyone know where they are in the HP site? |
|
|
Dave Hamblin
|
|
Nov 2, 2009 18:14:35 GMT
N/A: Question Author
|
|
|
Never mind, I found them, I am closing this thread as resolved. |
|
|
Dave Hamblin
|
|
Nov 2, 2009 18:15:08 GMT
Thread closed by author
|
|
|
Printable version
