Do I need to stop i/o when I upgrade SAN switch, and if yes then how ?

[Hanry Zhou]

I have StorageWorks 4/32 with version of 4.4.0b. Hosts has two HBA's and are connected two different redundant switches. I am planing to upgrade the version one at one time.

My question is, before I start to perform the upgrade, Do I need to stop i/o first through this switch, if I do, what steps I need to go through to stop the i/o?

Thanks

[Rob Leadbeater · Accepted solution]

Hi,

If you have redundant switches, and multipathing is working properly at the OS level and on the storage system(s), then you should just be able to upgrade the switches one at a time.

The switches will reboot during the upgrade process, and when that happens the I/O will failover to the redundant path.

If you're not confident that the multipathing is working correctly, then yes you should stop all I/O, which essentially means shutting down all the servers that are connected to the SAN.

Hope this helps,

Regards,

Rob

[Håkan Persson_1]

Firmware updates for the Brocade switches are non-disruptive (with exceptions for some older switches and FW3.x.x). So I disagree with Rob that i/o will failover.

[Sheldon Smith]

The key in Rob's response is "...multipathing is working properly at the OS level...".

For peace of mind, go to each server connected, and check it's multipathing software to ensure all paths are good. Depending on OS and multipathing software, you may be able to manually make the system _prefer_ a particular path, and also change the preferred path to verify all paths work. Before you start, read your multipathing software documentation, and find how to make a note about how the multipathing software is currently configured. That is, whether it is using some sort of load-balancing across the paths, or whether each LUN has a preferred path.

"...StorageWorks 4/32 with version of 4.4.0b". The current recommended supported version is 6.2.1. That's a LOT of firmware upgrades:
4.4.0b to 5.0.x to 5.2.x to 5.3.x to 6.0.0c_upgrade to 6.1.x to 6.2.1.


-----
Please rate these answers (and the 950-some other answers) by assigning points on a scale of 1-10. To assign points, simply login and click the dropdown points menu next to each reply. Use the following scale when rating a response:

N/A: The reply was a clarification to my original question
1-3: The answer didn't help answer my question, but thanks anyhow!
4-7: The answer helped with a portion of my question, but I still need help.
8-10: The answer has solved my problem completely! I'm a happy camper!

[Click on your name, you will be taken to your Profile. Down under "My question(s)" is a link to "Questions or topics with unassigned points". Click there.]

WHY SHOULD YOU DO THIS?
There are 3 reasons why we feel rating replies is such an important feature:

1.) Others have taken time to help you, so please give them credit for their help.
2.) Your rating will help your peers earn points toward their Forums status, and you will validate the quality of the solution you've received.
3.) Other readers will understand which answer best solved a problem, enriching the community knowledge being shared.

[Hanry Zhou]

Thanks all for your inputs.

now assuming redudance works as expected, I understand that I can work on upgrading on one switch, without losing connections between servers and XP storage. But, happend to these i/o data within this swtich while I am doing the upgrade, are all these i/o will be moved over to the other switch?

I know the fact as you guys just said, but, I am just not sure of in detail how it will out out.

Sheldon, I will do better to score points. Thanks,

[Rob Leadbeater]

Hi,

It's been a while since I've worked on a 4/32 but I'm pretty certain it will reboot towards the end of the upgrade process - when the new code gets loaded into memory.

At this point if there is any I/O going through the switch it should failover to the redundant path.

Hope that helps,

Cheers,

Rob