Try 'powermt check force' I'm not too confident that it will work, but worth a shot.
Something is holding the device open, my guess would be that its the queued IO outstanding... see the the powermt display output:
queued-IOs=1
Two possible solutions, 1) If this host is clustered, perhaps a cluster service is holding the device open, try stopping cluster services and 'powermt check'
2) Are you in a position to reboot the host? If not, for the next reboot ensure immediately afterwards you perform a 'powermt check' and 'powermt save'
powermt config did not help which we tried already.
these devices were under ASM diskgroup (Oracle CRS cluster / 10g) and disk group members were pointed to the new LUNs. (We moved a few LUNS to a different array)
I've had a similar issue before. I don't think I found anything except a reboot.
Also, if you have Solution Enabler installed on this host, try stopping all it's daemons and checking with 'symcfg list -semaphores', maybe something is using that device. Best of luck!
powermt display shows device path as UNKNOWN and path as DEAD on HPUX powermt check will not remove dead path and returns an error saying the device path is in use vgdisplay -v shows the dead path as the only physical volume in the volume group
New switch added to environment. Unplugged one port of service processor from old switch and plugged it into the new switch. This caused HPUX to generate new paths. Because the path was the only physical volume in the volume group, dead paths could not be removed from PowerPath as the volume manager was still using it.
Follow the steps below to resolve this issue:
1. Unmount all file systems in the volume group
2. Add one of the paths that is there (from the powermt display dev=all) using the vgextend command
3. Remove the dead path using the vgreduce command
cause of "unknown" is identified as the devices file (cxtydz) were removed by rmsf which is suppose to be done after "powermt check" clears the dead path.
The devices involved are from Clariion. So what is the equivalent for "symcfg list -semaphores" to list any locks .
Conor
341 Posts
0
January 20th, 2009 14:00
Something is holding the device open, my guess would be that its the queued IO outstanding... see the the powermt display output:
queued-IOs=1
Two possible solutions, 1) If this host is clustered, perhaps a cluster service is holding the device open, try stopping cluster services and 'powermt check'
2) Are you in a position to reboot the host? If not, for the next reboot ensure immediately afterwards you perform a 'powermt check' and 'powermt save'
rawstorage
419 Posts
1
January 20th, 2009 08:00
You may need to remove it from veritas first
Primus emc156649 deals with this
Message was edited by:
Paul Martin
SKT2
1.3K Posts
0
January 20th, 2009 08:00
these devices were under ASM diskgroup (Oracle CRS cluster / 10g) and disk group members were pointed to the new LUNs. (We moved a few LUNS to a different array)
dynamox
1 Rookie
1 Rookie
•
20.4K Posts
0
January 20th, 2009 08:00
SKT2
1.3K Posts
0
January 20th, 2009 10:00
dynamox
1 Rookie
1 Rookie
•
20.4K Posts
0
January 20th, 2009 10:00
MaximUnited
111 Posts
0
January 22nd, 2009 00:00
I don't think I found anything except a reboot.
Also, if you have Solution Enabler installed on this host, try stopping all it's daemons and checking with 'symcfg list -semaphores', maybe something is using that device.
Best of luck!
Conor
341 Posts
0
January 22nd, 2009 06:00
powermt display shows device path as UNKNOWN and path as DEAD on HPUX
powermt check will not remove dead path and returns an error saying the device path is in use
vgdisplay -v shows the dead path as the only physical volume in the volume group
New switch added to environment. Unplugged one port of service processor from old switch and plugged it into the new switch. This caused HPUX to generate new paths. Because the path was the only physical volume in the volume group, dead paths could not be removed from PowerPath as the volume manager was still using it.
Follow the steps below to resolve this issue:
1. Unmount all file systems in the volume group
2. Add one of the paths that is there (from the powermt display dev=all) using the vgextend command
3. Remove the dead path using the vgreduce command
4. Run powermt check to remove dead path
5. Mount file system again
SKT2
1.3K Posts
0
January 22nd, 2009 06:00
i dont have SE on this node right now.
Now the devices are showing as unknown instead of the respectice ctd numbers.
xxxx 600601609E8415002884C4347881DC11 [LUN 720] 48 1/0/8/1/0.97.76.19.0.1.7 unknown SP A0 active dead 0 1 - NOT found in lvmtab
Are we hitting any bug with PP? I think the "unknown" is NOT casued by the force option.
SKT2
1.3K Posts
0
January 22nd, 2009 07:00
SKT2
1.3K Posts
0
January 24th, 2009 11:00
The devices involved are from Clariion. So what is the equivalent for "symcfg list -semaphores" to list any locks .
SKT2
1.3K Posts
0
September 22nd, 2009 05:00