How to stop a LUN

[Gary L]

Hi

For some reasons, we want to stop the access between the HP-UX sever and SAN storage. We need Shutdown a LUN and startup/restore it after finish the upgrade.
I could use spmgr display see this LUN and could use strings /etc/fstab find out the volume group, LV and mount point. I want to know how to stop and restart a LUN path? spmgr delete / spmgr add ? or umount the mount point? or vgchange -a n /dev/vg03? ...

thank you very much?

[Sundar_7 · Accepted solution]

Gary,

Stop a Lun ? If you mean preventing the system from using the LUN then I would suggest unmount the filesystem. Ofcourse, that will only work if the LUN is being used for a filesystem.

If it is being used as raw LV/device, then you can make the LV unavailable using the lvchange or vgchange commands.

Sundar.

[Sandman!]

What are you trying to do that you need to stop all access to this LUN? For doing this you would need to identify all lvols that have extents on this LUN and then either unmount the filesystems on this LUN or lvchange the availability of all lvols to "n" so that no new processes can access them lvols.

~cheers

[Court Campbell]

Not sure I understand this either. But off the cuff, I would say just pull the fiber connections from the server. No more SAN connection.

[Gary L]

Hi Sundar and Sandman Court Campbell

Thanks for your fast reply.
As you know, we wanna do some upgrade and fix on storage, the task requirement wrote "stop the access to SAN (LUN ID: 60001fe1...), I think, maybe its mean was let me stop the connection between HP-UX and SAN. And restore it after fixing.
You means, I just vgchange -a n /dev/vg03 and umount the all the corresponding mount points could stop the connection, right?

[Gary L]

Hi Court Campbell

"pull the fiber connections from the server. No more SAN connection." What's your mean? would you please give me more details?
thanks.

[Sandman!]

>stop the access to SAN (LUN ID: 60001fe1...)

Well you could simply unmount all the lvols that have extents on this LUN or you could also deactivate the VG that houses this LUN but that would stop access to all file systems that don't have extents on this LUN. Run pvdisplay on this LUN's device file and post the output here.

~thanks

[Court Campbell]

Umm. Your server is connected to your storage device somehow. I am assuming you have 1 or 2 fiber HBAs that connect you to your SAN. If you pull the fiber connections from the back of the server you will no longer be connected to it.

NOTE: I am not saying it is the correct thing to do.

Also if you do chooose to do that, you would want to stop any processes etc that are using the disks in the SAN and umount the mount points.

[Gary L]

through command spmgr display I find the LUN ID devices is c8t0d0
> pvdisplay /dev/dsk/c8t0d0
--- Physical volumes ---
PV Name /dev/dsk/c8t0d0
VG Name /dev/vg03
PV Status available
Allocatable yes
VGDA 2
Cur LV 3
PE Size (Mbytes) 32
Total PE 17490
Free PE 1461
Allocated PE 16029
Stale PE 0
IO Timeout (Seconds) default
Autoswitch On

through command bdf I could find the corresponding file systems of vg03 is /f01 /f02

according as your suggestion, I could use umount f01 / f02 or/and vgchange -a n /dev/vg03 stop the access and connection between HP-UX and SAN, right?

[Sandman!]

Do both /f01 and /f02 have extents on /dev/dsk/c8t0d0? If so then un-mounting them will be alright otherwise unmount only the filesystem that has extents on /dev/dsk/c8t0d0 (/f01 or /f02 or both).

[Gary L]

Hi Court Campbell

You are funny guy, thank you very much for your suggestion and help.
Yes, our HP-UX servers connect to the SAN storage via two FC HBAs each.
But this HP-UX server be placed in our another city's computer room. Unplug or pull the FC of server back for stopping the access is a "good" idea, my boss will fire me at the same time. Just a joke! I could not do this.

[Sandman!]

Another thing to keep in mind is that if this LUN is part of a high availability cluster like MC/ServiceGuard then it would be good to disable pkg switching, locally and globally, so that the pkg does not fail over to another cluster node where it could be activated alongwith its filesystems /f01 and /f02.

~hope it helps

[Gary L]

Hi Sandman

vg03 have two PVs (one is the c8t0d0) and two LVs, each LV be created file system via "newfs" and "mount" those LVs to f01 and f02. f02 is the mount point of c8t0d0.
according as your method, I just umount f02 and/or lvchange -a n lv_c8t0d0, right?
thanks

[Sandman!]

>I just umount f02 and/or lvchange -a n lv_c8t0d0

Un-mounting /f02 would make the filesystem unavailable and kill all processes that are currently using it. On the other hand lvchange /dev/vg03/lvol?? would prevent all new procs from accessing this lvol/LUN but the currently running processes would not be affected and would continue to run.

~hope it helps

[Gary L]

Got it!
Thanks Sandman,
thank you very much for everyone here.

Have a great day