PV links

[SAM_24]

Hi,

If PV links is not configured on the system , will it cause the system to hang? Assume disk1 is primary path and disk2 is PV link not configured into the disk group.

Thanks.

[Michael Steele_2]

Is your system hung? Then:

control B
return, return
GSP>rs (* This will reboot the system from the firmware level *)

Could you elaborate about PVLinks? For PV links come like so: Disk1 = primary and Disk2 = Alternate. Use:

vgdisplay -v /dev/vg## | more

...to determine.

LVM errors are recorded in syslog.log.

Additionally, here are to basic PVlink commands. This will set your primary PVlink:

pvchange -s y /dev/dsk/cXtYdZ

-or-

This will AUTOMATICALLY switch to alternate PVLink:

pvchange -s n /dev/dsk/cXtYdZ

Finally, I've seen bad system performance with PVLinks, but many things can cause a system to hang. Need more info.

[John Poff]

Hi,

Not having a PV link shouldn't cause the system to hang in a normal LVM situation. The PV link is there to provide an alternate path in case the main path is lost. Now, I guess it could appear to cause a problem if the controller that the PV link uses is having problems, but it really wouldn't be the PV link itself causing the trouble.

Are you having trouble with a disk on your system?

JP

[Sridhar Bhaskarla]

Hi Raj,

If you do not have PV Link, then the path to disk subsystem will not have redundancy. So, any link failure can cause issues. System won't hang per say, but the processes that access the disks will wait for their IO to complete. If you have OS filesystems like /opt/,/var,/usr etc on the external disks, then the system will hang.

If the PV link is configured, then LVM can switch the link to the alternate path if there is a link failure.

-Sri

[S.K. Chan]

Like what have been mentioned .. not having an alternate path configured would not in theory hang up your system/os. Your system might appear to be hung if you try to access data from a disk or path that is down even then it would come back and timeout on you with an error. Having said that I've experience a single disk failure on a HVD10 (this has a daisy chain connection to another disk array - H12) which caused the whole bus to hang and the system as well. In my case it was due to some patches I'm missing. Also I had alternate paths set up. So part of the answer to you question is "it depends", for most part the answer is no.

[Eugeny Brychkov]

You can not configure disk into vg without giving to vg device file of disk. Thus if you have disk configured in vg then there's at least one link to the disk. If you have only one link to the disk in vg config then if this link fails vg may become inaccessible (if there's no properly configured mirror copy of this disk).
The answer to your qusetion is:
if you have disk in root vg (vg00) AND there's no mirror copy of this disk (configured with MirrorDisk/UX) AND there's only one path to this disk exists in vg configuration THEN when this disk fails (becomes inaccessible) system may crash/panic/hang.
That's why usually people doing mirroring of root disks, and these disks are seating on different SCSI busses. Example: N-class/L-class servers
Eugeny

[SAM_24]

Thank you all!!!
I got the answer from you.