PV and LV timeout values for rolling SAN Firmware update

[GTownes]

Hi All,

We are planning updating our SAN environment to the latest firmware. We have been advised to we can do rolling updates, but must set timeout values greater than 60 seconds on the applications, and OS. We are running wndows and HPUX in our environment.

With regards to the HPUX OS, would recommend changing the timeout on the PV or both PV and LV?

[Torsten.]

What is your "SAN"?

I assume you are talking about a fibre channel disk array.

[IBaltay]

Hi,

Intro
1. Selecting a period of low I/O activity is essential
2. checking the MPIO/native MPIO/PVlinks/Autopaths both path activity and accessibility to the array

3.For all your platforms
you should go to 60 (HP-UX/Windows standalone) or more seconds of 90-120 (HP-UX SG/MSCS clusters) seconds

[IBaltay]

The above applies only to EVA storage, XP storage microcode upgrade is fully online-nondisruptive as well as the SAN switches firmware upgrade

[IBaltay]

but the first prerequisite (which must not be forgotten) especially in the storage controller firmware ugprade is to upgrade all the hosts HBA BIOSes/firmwares/drivers/multipathing drivers at SPOCK:
http://h20272.www2.hp.com/

[GTownes]

Thanks for the replies guys.

Yes, I am referring to EVA's. EVA 8100 and 8000. The HPUX servers hba's are up to date with firmware. I am just trying to figure out if I need to set just the PV timeout or both PV and LV timeouts on the servers.

[IBaltay]

lvdisplay = "forever"
pvdisplay = 30s.
So the PV timeout value should be changed to 60s

[Torsten.]

For HP-UX the driver is much more important than the firmware, because the driver will replace the firmware while OS is running.

Now it depends - what version of hp-ux, and multipathing software in use?

[Torsten.]

This will help too


http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00795548/c00795548.pdf




HPStorageWorksEnterpriseVirtualArray
online ï¬ rmwareupgradebestpractices

[Uwe Zessin]

I would be *very* careful on these large boxes as the 60 seconds might not be enough:

c01815339 - HP StorageWorks Enterprise Virtual Array (EVA) - Online Firmware Upgrades May Take Longer Than Expected

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01815339

(the support people who are pressing the "GO!" button do know why they want to do offline upgrades...)

[GTownes]

This great information. I've changed my PV timeout to 180 sec. I guess one thing I need to get clear is, do I need to change the LV timeout to 180 or leave it "0" or should just the PV timeout be enough. According to the document provided it only shows changing the timeout for the PV.

[IBaltay · Accepted solution]

yes LV timeout can stay as is:

Without LV timeouts, the system continues to retry an I/O to a nonresponding disk, possibly forever. If the disk finally responds, the I/O successfully completes. However, in a case when the disk never responds, the I/O will never complete and then never return to the caller. Thus, the caller will be "hung" forever waiting for an I/O that will not complete. If an LV timeout is specified (via the lvchange command with the -t command option), I/O to a nonresponsive disk will also be retried, but only for a length of time that does not exceed the specified timeout value. If the disk fails to respond within that time, the system will return an I/O error to the caller; thus the caller will not "hang" longer than the specified timeout value.

[GTownes]

Thanks guys for your quick and helpful responses.

G

[skt_skt]

I have never heard of recommending LV time out change. But always keep pv time between 90-180 seconds depending on the disk/array.