[Linux-cluster] DLM Problem
jr
johannes.russek at io-consulting.net
Fri Feb 1 10:24:46 UTC 2008
Am Donnerstag, den 31.01.2008, 19:29 -0600 schrieb isplist at logicore.net:
> I've not looked into this yet so don't know what to edit or add.
>
> On one node, the drive is /dev/sda, but on some nodes, it'll be sdx because
> some of the machines already have SCSI drives in them. Not hard to see, just
> have to keep it in mind since I'm assuming each machine needs this file.
>
> I see a catch all ( sd*[!0-9] ) but don't know where I need to put say
> /dev/sda.
okay, i'll try to help as far as i can with my nonexistant udev rules.
the catchall is there because you *don't* know which scsi disk it is,
wether it is sda, sdb or whatever.
however, the DISK does have a unique serial number (even raid slices do)
or at least a WWN.
so you basically need to find out which unique number your unique disk
has (by using /sbin/scsi_id -p 0x83 -g -u -s /block/sda), tell that the
udev system and let it create a symlink you name whatever you like. this
way, every node get's a symlink with the same name for the same disk,
whatever the state of it's device enumeration is.
oh, any udev gurus.. i suppose udev uses page 0x83 rather then 0x80, is
that correct?
regards,
johannes
More information about the Linux-cluster
mailing list