vgimport/change hangs machine

[Pete Devlin]

We have a K580 6 way running 11 64 bit as part of a 2 node Service Guard Cluster. We keep coming accross an issue where, if disks on one side of a HASS are imported into any vg or created as new & vgchange is run the machine hangs or panics. We have rebuilt the machine several times now & have a new HSC card,SCSI cable, HASS backplane & even disks. This orginally showed it's head when the disks were configured as root mirrors - which worked but the machine became unbootable hence the rebuilds.
I've done a search & can't really find much about this. The partner machine is absolutely fine.

[harry d brown jr]

Check the SCSI addressing, make sure you don't have two disks with the same ID. Also, replace your scsi terminators - they do fail. You don't have the two dies crossed over do you?

live free or die
harry

[Pete Devlin]

We've changed the checked & chnaged (just for good measure!) the SCSI ID's & terminators. However we still have the same problem. We're now thinking it is almost certainly gonna be software related. I's weird that it happens even after system build.

[Sanjay_6 · Accepted solution]

Hi Pete,

Looks more likely as a patch issue. Are both your system at the same patch level. Another thing that i would suggest is replacing all the disks on that side of the hass and then try with addiing disks from that side one by one into the VG, then do a export and import. This will isolate a sigle disk, if that is giving the problem. I had a similar problem with one of my hass sitting between two 4 way k460's and i isolated the problem with one disk. I have kept that disk out of the VG and am using another set of disks for the time being, had a deadline to meet.

Hope this helps.

Regds.

[G. Vrijhoeven]

Hi,
This is a long shot but you never know.
The hanging of a node could be coused by the removal of the cluster lock disk on one node. Check cluster configuration on one node and check the cluster lock disk (vg).

Succes Gideon