how to unlock SAN disk?

[ogniemi]

I have a LUN shared between two systems (like in HACMP configuration but i it not HACMP installation in this case).

Let's say node01 having shared VG varied on crashes and I can't varyonvg on node02 due to lack disks. What command should I run on node02 to unlock the disk/LUN? Right now, I have to power on node01 and run there vayonvg/varyoffvg on node01 to be able varyon the VG on node02.

Is it allowed to use following binary from HACMP for that? :

/usr/es/sbin/cluster/events/utils/cl_fscsilunreset

Will it work?

Any ideas?

 

[khalidaaa]

let me get what you said, now the vg is varied-on on node01 and you want to vary it on node02 at the same time? As i understand theoritically, (never happened to me and i don't hope it happens ), that if you manage to vary on a vg that is already varied on to another vg then you will corrupt the data on that vg!

Is you vg enhanced-concurrent?

Why do you want to do that any way?

never tried that command (/usr/es/sbin/cluster/events/utils/cl_fscsilunreset) without hacmp so i don't know the effect to be honest!

Regards,
Khalid

 

[ogniemi]

no, I don't want to varyon the VG on obth nodes at the same time. But when a node having the shared VG online crashes (halts) and SAN disks were not unlocked they are still locked on DS4800 so I cant use them (varyonvg) on second node in that case.

I know what HACMP handles with this problem running cl_fscsilunreset command but I expect there is another way I can use having no HACMP installed. copying cl_fscsilunreset binary is the last I would like to do if no the other solution exists.

 

[khalidaaa]

Is the vg enhanced concurrent? i'm not sure if cl_fscsilunreset will work if it is not enhanced concurrent!

were you be able to varyoff the vg on node01 and vary it on on node02 in normal conditions?

I don't see why you don't want to use the HACMP in your case any way! But nice scinario

 

[p5wizard]

If you ever need to un-scsi-reserve a disk, any utility that can do that is fine. You could even write your own C code to do it (IMHO you need to execute an ioctl call that unlocks/resets the persistent SCSI reservation on a disk device).

Also see if the driver you installed for the SAN box has a utility to do that. E.g. SDD for ESS/DS6k/8k has a utility /usr/sbin/lquerypr that can be used for that.


HTH,

p5wizard

 

[DukeSSD]

the reservation is probably held by the ds4800 in lue of the node01, so without the node01 releasing the lock you would need to use the ds4800 controller / manager interface to force release the lock. sorry.

 

[ogniemi]

DukeSDD,
in HACMP env. cl_fscsilunreset seems to do the trick. when luns are locked by node01 which is not avaiable (crashed) luns must be unlocked to perform successful resource group takeover to node02

p5wizard,
thx for you suggestion.

 

[DukeSSD]

Worth knowing, my san is managed by "The SAN team" so i have (had) to ask them to release locks left over from down / dead non-ha (usually test / dev) systems, I'll load up the cl_fscsilunreset binary and do it myself in future, can't see any harm in that, thanks.