Disk Array Reports Missing Disk

[Ray Evans]

Our model 12 Array failed and in order to release the bus we had to pull a disk without downing it first.

Is there a method to tell the Array that the disk is gone and to forget about it?

[Bill McNAMARA_1]

Yes, there is, you can use mstm I believe to do this.

With arraymgr I believe it is possible also (with -Z option) , however, I don't have access to my 12H system today to verify..

mstm is possibly the easiest solution. - check the expert tools after selecting the 12H.


Bill

[Insu Kim]

See the below for detail.
I got this in a 12H manual and simply copied it for your reference.

The commands to down a disk can be either destructive, or non-destructive.
If the disk is returned to the array configuration,
this determines whether it will be seen as containing any valid data. If a destructive down has been performed, the array will
treat the disk as containing no valid data. If a non-destructive down was performed, the array will assume that all data on the disk is valid.

Downing a disk for non-destructive testing:
# arraycfg ???D b6 ???d ???v

To protect user data, the array will not allow you to down a disk
if doing so would result in loss of data redundancy or data availability. You can override this protection by specifying the
appropriate option:
-R allows downing with loss of redundancy, no loss of availability
-Z allows downing with loss of redundancy and/or availability

Downing a disk for destructive testing, allow loss of redundancy:
# arraycfg ???D a3 ???d ???R

A downed disk can be returned to the disk array configuration by manually including it using either the Control Panel, or the
ARMServer utilities:
# arraycfg ???D a3 ???a

[Nico van Royen]

According to its man page, you can use the arraycfg command in /opt/hparry/bin like:

# arraycfg -D -Z
(i.e. arraycfg -D A1 -Z if A1 is that disk)


options for these:
-D:
Perform a physical configuration operation on the disk installed in the cabinet slot identified by slot. Slot numbers must be of the form An or Bn, where A or B correspond to a cabinet column, and n corresponds to a shelf position (1-6).

-Z:
Allows the disk to be downed even if data unavailability would occur. If -Z is not included, the command to down a disk will
fail if it would result in data unavailability.
WARNING: Specifying the -Z option could cause
the disk array to become unavailable.

[Ray Evans]

All of these suggestions would work, but do not fit our situation.

Maybe I should have provided more detail:

Our file systems on our array suddenly stopped responding to IO requests. Diagnostics reported disks devices on the array as no longer accessible. from remote, I could not see what was wrong, once I came on site, the array had both controllers and control panel status lamps blinking about twice a second. Disk A3 was lit continously. Speaking with support center we determined that A3 had crashed hard and was locking up the internal SCSI bus. As we could not get to the array via the host nor did it respond to the commands from the control panel, we decided to pull the disk. Once pulled, the system started processing commands that had been in queue (to include a shutdown -r) and the array began a rebuild process. The rebuilds would never succesfully complete. The array reported disk A3 as missing because we could not get it to accept a down command.

So this is the question:: After removing a disk without first downing it, can you somehow tell the array to forget that disk?

[Nico van Royen]

We have had a similair problem on some occosians.
On our model 12, the array would 'hang' (both controllers with so much activity that they would not accept any other data from remote).. but AFTER we manually pulled a disk that was reported as failed by the array, or sometimes after inserting a new disk.
Our solution for the 'hang' of the array was to pull one of the controllers, wait a few seconds and place it back in. This however is not your solution at this time. I would try to re-insert the crashed disk (or better yet, a new one).. the array will start to initiate the rebuild process. After that has completed (or failed) you can then down the disk and pull it again after its downed.

As as sidenote, you might look into the firmware revision of your array and the disks inside them and perhaps update them if they are out of date to prevent future hangups.

- Nico