[gpfsug-discuss] NSD device type

Arif Ali mail at arif-ali.co.uk
Fri Sep 14 10:38:15 BST 2012


So a reply on this.

We bit the bullet and went ahead, and it worked a treat, literally
took 5 seconds to do.

On another note, something that would be useful for others. We all
know that without rdac the nsd node would moan for ~15 mins depending
the number of disks with I/O errors for any machines that are fibre
attached. One way to supress that would be to add the following kernel
parameter

rdloaddriver=scsi_dh_rdac

This saved us many hours on installing these machines. I hope it will
be useful for others.

For the purpose of google, so that they can find the info easily,
adding an excerpt of the error message below

sd 6:0:1:8: [sdal] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 6:0:1:8: [sdal] Sense Key : Illegal Request [current]
sd 6:0:1:8: [sdal] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 6:0:1:8: [sdal] CDB: Read(10): 28 00 00 00 00 40 00 00 08 00
end_request: I/O error, dev sdal, sector 64
sd 5:0:11:0: [sddc] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 5:0:11:0: [sddc] Sense Key : Illegal Request [current]
sd 5:0:11:0: [sddc] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 5:0:11:0: [sddc] CDB: Read(10): 28 00 00 00 00 40 00 00 08 00
end_request: I/O error, dev sddc, sector 64
sd 5:0:11:8: [sddk] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 5:0:11:8: [sddk] Sense Key : Illegal Request [current]
sd 5:0:11:8: [sddk] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 5:0:11:8: [sddk] CDB: Read(10): 28 00 00 00 00 40 00 00 08 00
end_request: I/O error, dev sddk, sector 64
sd 6:0:6:4: [sdhb] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 6:0:6:4: [sdhb] Sense Key : Illegal Request [current]
sd 6:0:6:4: [sdhb] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 6:0:6:4: [sdhb] CDB: Read(10): 28 00 00 00 00 00 00 00 08 00
end_request: I/O error, dev sdhb, sector 0
sd 6:0:6:8: [sdhf] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 6:0:6:8: [sdhf] Sense Key : Illegal Request [current]
sd 6:0:6:8: [sdhf] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 6:0:6:8: [sdhf] CDB: Read(10): 28 00 00 00 00 80 00 00 08 00
end_request: I/O error, dev sdhf, sector 128
sd 5:0:16:8: [sdex] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 5:0:16:8: [sdex] Sense Key : Illegal Request [current]
sd 5:0:16:8: [sdex] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 5:0:16:8: [sdex] CDB: Read(10): 28 00 00 00 00 00 00 00 08 00
end_request: I/O error, dev sdex, sector 0
sd 6:0:6:6: [sdhd] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 6:0:6:6: [sdhd] Sense Key : Illegal Request [current]
sd 6:0:6:6: [sdhd] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 6:0:6:6: [sdhd] CDB: Read(10): 28 00 00 00 00 40 00 00 08 00
end_request: I/O error, dev sdhd, sector 64
sd 6:0:6:2: [sdgz] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 6:0:6:2: [sdgz] Sense Key : Illegal Request [current]
sd 6:0:6:2: [sdgz] <<vendor>> ASC=0x94 ASCQ=0x1ASC=0x94 ASCQ=0x1
sd 6:0:6:2: [sdgz] CDB: Read(10): 28 00 00 00 00 40 00 00 08 00
end_request: I/O error, dev sdgz, sector 64

--
Arif Ali

catch me on freenode IRC, username: arif-ali


On 13 September 2012 19:50, Arif Ali <mail at arif-ali.co.uk> wrote:
> All,
>
> We are moving from rdac to multipath on one our customer
> installations, and the nsd type of the disk is generic, but when
> multipath is required, we need the device type to be changed to dmm.
>treat
> We know we can change it using "mmchconfig updateNsdType=$PWD/nsdchange"
>
> where file nsdchange has the following input
>
> <nsdname> dmm
>
> Would this have any impact on the disks themselves, or does this only
> change /var/mmfs/gen/mmsdrfs.
>
> Any clues on this would be great
>
> --
> Arif Ali
>
> catch me on freenode IRC, username: arif-ali



More information about the gpfsug-discuss mailing list