[dm-devel] NIC and HBA based multipathing

[Mike Anderson]

Mike Christie <michaelc at cs.wisc.edu> wrote:
> Scott Moseman wrote:
> >On Nov 13, 2007 10:52 AM, Scott Moseman <scmoseman at gmail.com> wrote:
> >>># ls -l /dev/sd* | grep -v sda
> >>>brw-------  1 root root 8, 16 Nov  7 14:43 /dev/sdb
> >>>brw-rw----  1 root disk 8, 17 Nov  7 14:43 /dev/sdb1
> >>>brw-rw----  1 root disk 8, 33 Nov  7 14:45 /dev/sdc1
> >>I can use 'mknod sdc b 8 32' to generate a new /dev/sdc, which I can
> >>fdisk and the data looks good, but once I reboot the /dev/sdc device
> >>is once again removed (or not re-created, whichever the case may be).
> >>
> >
> >Does anyone know why the /dev/sdc device might not get created on
> >boot?  Is this something that the iSCSI initiator should handle?  The
> 
> The iscsi initiator is not responsible for creating device nodes. If you 
> see this:
> 
> SCSI device sdb: 8417280 512-byte hdwr sectors (4310 MB)
> SCSI device sdb: drive cache: write through
>  sdb: sdb1
> Attached scsi disk sdb at scsi1, channel 0, id 1, lun 0
> 
> then the initiator and scsi layer have done everything they can.
> 
> >operating system?
> 
> It seems like a udev problem, but I do not know for sure.

I would agree that we should look to see what udev is doing when it
receives the events.

Since I believe this is failing on a RHEL4 sytem we cannot use
udevmonitor. If you unload the iSCSI module, edit /etc/udev/udev.conf to
set udev_log="yes", and then load the module again you should get some info
in /var/log/message on what udev is doing.

-andmike
--
Michael Anderson
andmike at linux.vnet.ibm.com