SCSI: Async write error

[Manminder Takhar]

Hi

I am getting the following error on a shared disk in a cluster.
- the cluster in HPUX 11i
- the disk is shared between 2 parisc L class servers
- during configuration the process went fine
- but when i try to test the disk it gave lot of errors
- strangely if i connect to only 1 server each time, i could cp large files on disk

Any help would be useful, patch? configuration?


Mar 1 17:47:11 hpsgnak vmunix: blkno: 20360, sectno: 40720, offset: 20848640, bcount: 8192.
Mar 1 17:47:11 hpsgnak vmunix: blkno: 20352, sectno: 40704, offset: 20840448, bcount: 8192.
Mar 1 17:47:12 hpsgnak vmunix: SCSI: Async write error -- dev: b 31 0x070000, errno: 126, resid: 40960,
Mar 1 17:47:11 hpsgnak vmunix: SCSI: Async write error -- dev: b 31 0x070000, errno: 126, resid: 8192,
Mar 1 17:47:12 hpsgnak above message repeats 95 times
Mar 1 17:47:12 hpsgnak vmunix: blkno: 19264, sectno: 38528, offset: 19726336, bcount: 40960.
Mar 1 17:47:12 hpsgnak vmunix: SCSI: Async write error -- dev: b 31 0x070000, errno: 126, resid: 65536,
Mar 1 17:47:12 hpsgnak vmunix: ...
Mar 1 17:47:12 hpsgnak vmunix: LVM: VG 64 0x020000: Reestablished quorum.
Mar 1 17:47:12 hpsgnak vmunix:
Mar 1 17:47:12 hpsgnak above message repeats 233 times
Mar 1 17:47:12 hpsgnak vmunix: LVM: VG 64 0x020000: PVLink 31 0x070000 Recovered.


Thanks
Manminder

[Andrew Rutter]

hi, probably more configuration than hardware problem

when you configured it initialy did you set the scsi controllers to different scsi id's? one should be 6 and the other 7.

then you need to make sure that the disk id's dont clash with these.

Are the disks shown in ioscan on both systems?

Also compare how the pv's/vg's are dispalyed on each system.

Andy

[Manminder Takhar]

Hi,

Please note:
1. i have another shared disk in the same disk enclosure and have no problem with it
2. so the scsi controller must be ok since other disk access has no issue
3. all the scsi ids are ok controller is 6 and 7 and the disks are different.

Another input is I this is a new disk to replace a previously existing disk. It was smaller so I had to put in a new one. Data on the disk was not a issue. So I did vgexport to remove the VG, and then recreated everthing. Do u see any issue with this?

I dont mind redoing the disk setup? What are the right steps to clear the disk i.e. the removing the disk file etc.

thanks
Manminder

[Kent Ostby]

Errno 126 means EPOWERF or POWERFAIL.

It's not always a true powerfail, but could be.

To see what device it is, use the dev number listed above and do the following:

ll /dev/dsk | grep 0x070000

Best regards,

Oz

[sathish kannan]

Hi there,

Do you get the same message on both servers? Have you checked accesing from the another node? Some time it may be due to a SCSI reservation against the disk on the Storage. I am sure MCSG shouldn't be using any reservation but reservation muight have been left on the Storage if these disk were accesed earlier by other host.

Regards
sathish

[A. Clay Stephenson]

You are not allowed to have read-write access to multiple hosts for the same LVOL unless you are only using raw i/o. An exception to this is if your are running 11.23 and are running the Cluster File System (CFS).