slow vgimport and vgchange

[Terence Lee_1]

I have a new rp7420 server using HDS as data storage. When I try to do a vgimport or vgchange for the disks on the HDS, it is very slow. Here's the results:

#> timex vgimport -m vg38.map -s /dev/vg38
real 8:15.50
user 0.01
sys 0.12

#> timex vgchange -a y vg01
Activated volume group
Volume group "vg01" has been successfully changed.
real 13.50
user 0.00
sys 0.01

An initial suspect is due to hardware on the switch, HBA card or storage. However, I hope to get some useful advice here.

[Jose M. del Rio]

Hi, Terence:
not an answer but a question: we have just moved from a rp5470 to a new rp8420 (same Clariion). Activating volume groups lasts several minutes while in our previous rp5470 was very fast.
Did you find and answer to your issue?
Thanks.
Regards.

[Alzhy]

What kind of HDS array and do you use dual paths to each LUN (hence PVLINKS).. Or are you using some kind of active/active Pathing software.?

Have you considered moving to VxVM(even the free base version)? VxVM works best with HDS, XP's and EVAs as well as any other active/active arrays.

Your LVM vgimport/vgchange dilemma may very well be relate to how you the HDS LUNS are presented to your rp7420. Or you're probbaly missing an important LVM patch.. serach the patch databse for anything LVM related and see if you have them on your system.

[Devender Khatana]

Hi,

More likely it is a patch issue only. What is the performance normally otherwise ? If it is OK try installing PHCO_30698 which is latest LVM commands commulaive patch.

http://www6.itrc.hp.com/service/patch/patchDetail.do?BC=patch.breadcrumb.main|patch.breadcrumb.search|&patchid=PHCO_30698&context=hpux:800:11:11

HTH,
Devender

[Bharat Katkar]

Hi,

In addition could you check for proper connections done and look for any cabling issues. Also Check for proper terminations as well.

Hope that helps.
Regards,

[Tim D Fulford]

You did the import using the serial id of the disks (-s switch) this does mean it will query ALL the disks devices attached to it. I'm guessing that HDS probably servers up quite a few LUNs. So the OS is actively going through all the LUNs checking if its serial ID matches the volume group you have. This could be why it is taking so long...

Regards

Tim

[Jose M. del Rio]

Thanks for your suggestions:
- it is a Clariion CX700 System
- no pvlinks: using EMC PowerPath at the server
- PHCO_30698 is installed
- checked cabling (FC, not SCSI, so no terminators). If the issue persists, I will try with different fiber cables and switch ports. There are no errors in the Clariion SPâ s event logs nor in the switchesâ logs.
- Tim: I donâ t understand very well you comment. Iâ m issuing a â vgchange -a eâ command inside the control script (*.cntl) of a ServiceGuard package.

Iâ ll try to find the simplest test case: Iâ ll create a new LUN, new â standaloneâ VG (not included in a ServiceGuard package) and Iâ ll iss

[Jose M. del Rio]

Sorry for the formtatting problems.
Here I go again:

Thanks for your suggestions:
- it is a Clariion CX700 System
- no pvlinks: using EMC PowerPath at the server
- PHCO_30698 is installed
- checked cabling (FC, not SCSI, so no terminators). If the issue persists, I will try with different fiber cables and switch ports. There are no errors in the Clariion SPâ s event logs nor in the switchesâ logs.
- Tim: I donâ t understand very well you comment. Iâ m issuing a â vgchange -a eâ command inside the control script (*.cntl) of a ServiceGuard package.

Iâ ll try to find the simplest test case: Iâ ll create a new LUN, new â standaloneâ VG (not included in a ServiceGuard package) and Iâ ll iss

[Terence Lee_1]

We have resolved the problem by changing the kernel parameter, default_disk_ir from 0 to 1. It seems that setting this parameter to 1 for particular range of HDS storage will cause the problem. Our vendor engineer happened to found a similar case overseas and recommended the solution.

[Tim D Fulford]

Hi

I read that you fixed the problem. What I was saying was the vgimport may have taken a long time beca=use you did

vgimport -m -s /dev/

This will scan ALL the disks on your system trying to match the serial id/vgid in the top of your mapfile with the serial id or vgid on each disks. So if you have say 100 LUNs on your Clariion, then it will scan all 100, down every alternate link. Once it has identified all the disks belinging to the volume group it will then import them.

A quicker way is to explicitly name the devices e.g.

vgimport -m /dev/ /dev/dsk/c10t0d0 /dev/dsk/c10t0d1 .. etc ..

Regards

Tim