Most vxdisk (and vxdiskadd and vxdiskadm) commands are getting me
the error message "Device path not valid". vxdg says the disk is
unusable. vxva tries to invoke vxdisksetup and it says that the
disk is still associated. I've been through the user's guide,
the admin guide, and the man pages, but haven't found anything
useful or any information on that error message.
Any ideas on how to fix this without disrupting access to my
volumes?
I'm running Solaris 2.6 5/98 and VxVM 2.5.
Thanks...
Michael Cerrato
m...@cise.ufl.edu
CISE Department, University of Florida
[..]
> Most vxdisk (and vxdiskadd and vxdiskadm) commands are getting me
> the error message "Device path not valid". vxdg says the disk is
> unusable. vxva tries to invoke vxdisksetup and it says that the
> disk is still associated. I've been through the user's guide,
> the admin guide, and the man pages, but haven't found anything
> useful or any information on that error message.
What does "vxdisk list" and "vxprint -Aht" say? If it says that the plex is
in need of syncing then use "vxrecover -g diskgroup -sb" to bring them
online.
(Check the arguments for the above commands so they won't do anything
dangerous under Solaris. They work fine in LSM/Digital UNIX (which is some
version of Veritas - or pretty close anyway))
>
> Any ideas on how to fix this without disrupting access to my
> volumes?
If you indeed can run vxrecover then it should run in the background if you
give it the -b option. -s tells it to start the volumes afterwards. But
check with the man page so it's the same for Veritas/Solaris.
>
> I'm running Solaris 2.6 5/98 and VxVM 2.5.
>
> Thanks...
>
> Michael Cerrato
> m...@cise.ufl.edu
> CISE Department, University of Florida
>
>
--
Jorgen Pehrson (HP 9000/380, VAX2000, DECstation 5000/200 (NetBSD 1.3))
j...@spektr.eu.org (PDP11/73, PDP11/83, PDP11/83 (2.11BSD)), Intergraph 200
spektr.eu.org/~jp/ MicroVAX 3100 (NetBSD 1.3), VAXstation 4000/90 (VAX/VMS)
Michael Cerrato wrote:
> I've got two Sun A5000 (Enterprise Network Arrays) each hooked to
> two different machines. I've set up Veritas VxVM to mirror my
> file systems between the two arrays. I wanted to test that
> VxVM was actually working so I simulated a disk failure by
> turning off one of the disks in one of the cabinets. When I
> tried to write to a the volume, VxVM took that disk and plex
> offline. Now I've turned that disk back on, and I can't seem to
> convince VxVM that's it's no longer dead. Ideally, I'd like to
> be able to tell VxVM that the disk is now alive, but stale, but
> at this point I'd settle for VxVM allowing me to reinitialize the
> disk without disrupting the other disks in the plex.
>
> Most vxdisk (and vxdiskadd and vxdiskadm) commands are getting me
> the error message "Device path not valid". vxdg says the disk is
> unusable. vxva tries to invoke vxdisksetup and it says that the
> disk is still associated. I've been through the user's guide,
> the admin guide, and the man pages, but haven't found anything
> useful or any information on that error message.
>
> Any ideas on how to fix this without disrupting access to my
> volumes?
>
> I'm running Solaris 2.6 5/98 and VxVM 2.5.
>
> Thanks...
>
> Michael Cerrato
> m...@cise.ufl.edu
> CISE Department, University of Florida
I assume you have already tried vxdiskadm options 3,4,5, 11.
You might try 'vxdctl enable' to go out and rescan for any new
drives. Volume Manager may be out of sync with what devices (drives)
are actually out there and vxdctl enable can help resync VM.
Just a suggestion....
Dave
Something like (and I'm simulating this because I've already
reinitialized the disks):
coast# vxdisk list
DEVICE TYPE DISK GROUP STATUS
c1t48d0s2 sliced - - error
- - beach02 cisedg failed: was
c1t48d0s2
> If you indeed can run vxrecover then it should run in the background if you
> give it the -b option. -s tells it to start the volumes afterwards. But
> check with the man page so it's the same for Veritas/Solaris.
When I try to use vxplex to re-enable the plex (it gets disabled
when the disk goes away), it says that it can't be enabled
because the disk beach02 isn't associated with any physical
media. I'm assuming vxrecover will tell me the same thing, but I
haven't actually tried it.
"David E. Myers" <David...@East.Sun.COM> emailed me and
suggested that I try a "vxdctl enable" and then reattach the
stale plexes, and if that didn't work then use "luxadm
insert_device" (used to get the system setup while hot inserting
new disks). The vxdctl still left me with the disk not
associated with any media. The luxadm command did clear the
error condition as far as VxVM is concerned, and vxdisk stopped
reporting "Device path not valid." At that point I did something
like:
vxdisk -f init c1t48d0
vxdg -g cisedg adddisk blah02=c1t48d0
vxdg repldisk beach02=blah02
vxplex att homes homes-02
This causes the plex to be remirrored. That's fine if the disk
has actually failed and been replaced, but what I really want is
some way to tell VxVM to reassociate the disk with the media
without reinitializing the media. For example, say the cabinet
fails or I accidentally trip over the cabinet power cord.
Nothing has happened to the physical disks, but they will all be
disassociated from the VxVM disks. I'd have to reinitialize all
of them and remirror all my striped plexes. On the other hand,
when I actually did accidentally turn off an external SCSI
cabinet whose disks had a DiskSuite stripe, all I had to do was
reboot (the computer freaked out when an entire SCSI bus went
away). This stripe wasn't mirrored--if I had been using VxVM
there, I would have had to restore the entire file system from
tape.
Michael Cerrato
m...@cise.ufl.edu