PV link behavior on failover problem/question

[Unix Administrator_5]

Problem with PV Link failover
We have Adv JFS and Mirror/ux software intalled on our system (hp-ux 11.0).

We have an L1000 with 2 A5158A fc adapters connected via a switch to an MTI
D100 disk array.

We have two paths to the data C4t5d0 and C5t5d0. We have atlernate links set
up. Whenever we pull the cable to the primary link's hba, the fail over occurs
as it should be we get scsi errors until the path is returned.

Is this normal behavoir? If so is it safe.


We can recreate this on other systems.

Any help would be appreciated.

[melvyn burnard]

Is it normal?
Yes
Is it safe?
Define "safe". It won't hurt you, but you will be seeing system overhead while the system tries to reach the discs down the fc link.

[Ashwani Kashyap]

Yes it is a normal behaviour of LVM . You will keep getting scsi errors until your primary path is returned because LVM still knows that there are LUNS attached to that path ,which now it consider it as down .

I don't know what is safe but now your are running your LUNS on one path and the SCSI errors will keep reminding you to bring the system back to dual path .

[Jeff Schussele]

Yes, it's perfectly normal. It should also be noted that you'll have to return the alt links to the VG with

vgchange -a y /dev/vg_name

after the link is back up or you'll still have only the one link.

Rgds,
Jeff

[Jeff Schussele]

I should also add that you may need an
ioscan -fnC disk
prior to the vgchange so the system will see the HW paths once again.

Rgds,
Jeff