Mirror-UX

[Ashwani Kashyap]

I have a HP-9000 , K-410 system running HP-UX 10.20.I have a jamaica box connected with 8 Nine GB disks in it, 4 on side A and 4 on side B. THe 4 disks on side A are connected to a SCSI controller one one bus and the 4 disks on side B are connected to another controller in a different bus. These two controllers are located in the same physical slot but have altogether different hardware addresses.

LV's on the Disks on side A are mirrored to disks on side B using mirror-ux.

Now last week one of the disks on side A failed but its mirror copy never took over and the logical volumes on that disk never became available.Putting it in other way mirror-ux simply failed.

Infact I got strange messages in the system log indiacting a SCSI controller failure , but the rest of the disks on side A were still working fine.

THe HP Engineer had to replace the failed disk and the controller card as well.But interestingly we were able to build the filesystems from the mirror copy without any difficulty after replacing the disks.
So why didn't in the first place the mirror copies took over when the disk failed??

Could any one help me understand what happened with my system? Should mirror ux be relied ? And what should I do , for this not to happen again in future?

Thanks

Ash.

[Alan Riggs]

Could you post the message about the SCSI controller error? Also, is there any chance you looked at/saved an lvdisplay of the mirrored lvol from the time it was unavailable?

[Ashwani Kashyap]

Hi allan,

I am sorry ,I do not have that file to post the SCSI controller message.But I do remember several lines saying SCSI BUS RESET followed by several Hexadeciaml Numbers.

I tried to run lvdisplay, ioscan ,vgdisplay and several other diagnostic command but they wouldn't run.THey would simply hang.IT seemed like system had just stopped responding to anything.I could just do a more on the syslog file and that's how I found out about the scsi error.

[James R. Ferguson · Accepted solution]

Ash:

Interesting. Document #KBRC00001157 would suggest that failures are rare, or at least limited as you would expect. I can't find the reference to the Mirrored Disk/XL Reference Manual. From KBRC00001157:

Mirrored Disc and Nonrecoverable Conditions

The system will not be able to recover from the following error conditions:
o Failure of both drive partners.
o Failure of the source drive during the repair operation.
o Use of the SUSPENDMIRROR command on an inconsistent drive.
o Software errors.
o Direct modification of disc using debug.

For further information refer to the Mirrored Disk/XL Reference Manual (part number 30349-90003), page 4-17 (Disk Failure)

Hope this helps some.

...JRF...

[John Palmer]

A failed SCSI card is a rare event in my experience.

What I guess happened is that the combination of SCSI controller and disk kept timing out the bus and causing resets. These events take a relatively long time and if they were occurring for most (if not all) disk transfers then the system would appear to 'hang'.

Most disk failures result in the disk appearing to 'die' completely so that the failure doesn't affect any other devices on the bus. HP-UX continues quite happily in these circumstances provided the disk volumes are mirrored.

In this case I think you were unlucky.