Before stating the actual issue, how do I modify the configuration files to
remove a domain without xen loading. Basically I screwed up one of the domains
disk preventing xen from loading
The error I''m getting at bootup is
svc:/system/xvm/store:default: Method "/lib/svc/method/xenstored
start" failed with exit status 96
system/xvm/store:default misconfigured: transitioned to maintenance (see
''svcs -xv'' for details)
svcs -xv
svc:/system/xvm/store:default (Hypervisor Store Daemon)
State: maintenance since November 9, 2009 4:48:28 PM EST
Reason: Start method exited with $SMF_EXIT_ERR_CONFIG.
See: http://sun.com/msg/SMF-8000-KS
See: man -M /usr/share/man -s 1M xenstored
See: /var/svc/log/system-xvm-store:default.log
Impact: 4 dependent services are not running:
svc:/system/xvm/xend:default
svc:/system/xvm/domains:default
svc:/system/xvm/console:default
svc:/system/xvm/virtd:default
The domain was a fresh install - I''m more then happy to nuke it and/or
rollback xen itself (and reinstall it) but have no idea how to do either (most
of the destroy methods rely on xen actually working)
How I got into the mess :D
1) I''m using snv_126 and installed xen.
2) I was able to do the install / reboot in dom0 successfully
3) i then proceeded to install a centos domu which was also successful.
(I did this with a zvol at /tank/vm/centos54/disk0. I first did a mkdir
/tank/vm/centos54 before creating the vm itself with the zvol disk parameter)
4) After everything was working satisfactorily I did a reboot (to get rid of the
nfs mounted centos dvd) and everything went to hell..
5) Initial error = /tank/vm is not empty - cannot mount.
6) I tried removing /tank/vm/centos but was unable to (device busy).
7) I then succeeded in doing a mv which was successful
8) On reboot I got a large number of out of swap space errors
9) I rebooted into the pre-xen environment
10) I rebooted into the xen environment again which booted but gave the error
mentioned in the beginning of the post
i''m presuming with the mv I messed up where its looking for disks and
xen before loading checks if all the disks are around and fails at that point
(pure conjecture on my part).
Any advice on what config files to modify and how to eliminate the zfs zvol to
get xen to a pristine state
or any other advice to get a working system again?
--
This message posted from opensolaris.org
Muhammed Syyid wrote:> Before stating the actual issue, how do I modify the configuration files to remove a domain without xen loading. Basically I screwed up one of the domains disk preventing xen from loading > The error I''m getting at bootup is > svc:/system/xvm/store:default: Method "/lib/svc/method/xenstored start" failed with exit status 96 > system/xvm/store:default misconfigured: transitioned to maintenance (see ''svcs -xv'' for details) > > svcs -xv > svc:/system/xvm/store:default (Hypervisor Store Daemon) > State: maintenance since November 9, 2009 4:48:28 PM EST > Reason: Start method exited with $SMF_EXIT_ERR_CONFIG. > See: http://sun.com/msg/SMF-8000-KS > See: man -M /usr/share/man -s 1M xenstored > See: /var/svc/log/system-xvm-store:default.log > Impact: 4 dependent services are not running: > svc:/system/xvm/xend:default > svc:/system/xvm/domains:default > svc:/system/xvm/console:default > svc:/system/xvm/virtd:defaultHmm, did you boot back into metal? Does a "uname -a" show i86pc or i86xpv?> The domain was a fresh install - I''m more then happy to nuke> it and/or rollback xen itself (and reinstall it) but have no > idea how to do either (most of the destroy methods rely on > xen actually working)> > How I got into the mess :D > 1) I''m using snv_126 and installed xen. > 2) I was able to do the install / reboot in dom0 successfully > 3) i then proceeded to install a centos domu which was also successful. > (I did this with a zvol at /tank/vm/centos54/disk0. I first did a> mkdir /tank/vm/centos54 before creating the vm itself with the zvol disk parameter> 4) After everything was working satisfactorily I did a reboot (to get rid of the nfs mounted centos dvd) and everything went to hell..Rebooted dom0 or the domU?> 5) Initial error = /tank/vm is not empty - cannot mount. > 6) I tried removing /tank/vm/centos but was unable to (device busy).what''s the output of zfs list? What your going to want to do is zfs umount -f /tank/vm, then remove the directory, then zfs mount it.. Be very careful and make sure you really unmounted zfs before you start removing things.> 7) I then succeeded in doing a mv which was successful > 8) On reboot I got a large number of out of swap space errorsexample of one?> 9) I rebooted into the pre-xen environment > 10) I rebooted into the xen environment again which booted but> gave the error mentioned in the beginning of the post> > i''m presuming with the mv I messed up where its looking for> disks and xen before loading checks if all the disks are around > and fails at that point (pure conjecture on my part). No... If xenstore isn''t working, you are never touching anything guest related.> Any advice on what config files to modify and how to eliminate> the zfs zvol to get xen to a pristine state> or any other advice to get a working system again?Your system is is fine... It doesn''t sound like there is any permanent damage. It should be easy to recover.. First thing to figure out is why xenstore isn''t starting. I''m guessing your booting into metal and not into the hypervisor. MRJ
> Hmm, did you boot back into metal? Does a "uname -a" show i86pc or i86xpv?It''s showing i86pc - the hypervisor isn''t running>> 4) After everything was working satisfactorily I did a reboot (to get rid of the nfs mounted centos dvd) and everything went to hell.. > Rebooted dom0 or the domU?Rebooted the entire machine (which would be dom0 / solaris)>> 6) I tried removing /tank/vm/centos but was unable to (device busy). >what''s the output of zfs list? What your going to want to do >is zfs umount -f /tank/vm, then remove the directory, then zfs >mount it.. Be very careful and make sure you really unmounted >zfs before you start removing things.Zfs List output (removed rpool lines) NAME USED AVAIL REFER MO tank 3.72T 1.62T 3.65T /tank tank/vm 70.0G 1.62T 32.9K /tank/vm tank/vm/centos54 70.0G 1.62T 31.4K /tank/vm/centos54 tank/vm/centos54/disk0 70G 1.68T 5.13G ->> 8) On reboot I got a large number of out of swap space errors >example of one?Warning: no swap space to grow stack for pid 17 ipv4 FWIW These errors cleared up when I did a boot into the physical pre-xen machine and rebooted to bare-metal again As to why xen isn''t loading I couldn''t find any error messages beyond that it wasn''t loading (no specifics). Any logs that I could look into? I checked the logs in /var/svc/logs/system-xvm-* I''m trying to remove the directory and recreate it / will report back Thanks for all your help :) -- This message posted from opensolaris.org
I unmounted /tank/vm/centos54 /tank/vm I then removed the centos54 folder and moved the old one back and am getting the same error again Mounting ZFS filesystem (1/11) cannot mount /tank/vm: directory is not empty svc:/system/filesystem/local:default: WARNING: /usr/sbin/zfs mount -a failed: exit status 1 When I initially created the vm I created /tank/vm folder /tank/vm/centos54 folder and then created the domain using the following disk parameter --disk path=/tank/vm/centos54/disk0,size=70,driver=phy,subdriver=zvol I presume something got confused between path and zvols? -- This message posted from opensolaris.org
removing both folders /tank/vm/centos54 /tank/vm and rebooting got both the hypervisor up and centos itself running. Thanks for all your help Another question - I was planning on running multiple domU/VMs, I was planning on using a disk/path command as follows --disk path=/tank/vm/ubuntu910/disk0,size=70,driver=phy,subdriver=zvol Would that work or am I at risk of reverting to the failure mode. Also any insight on why the existing paths (that were created before the domU disk /zvol was) caused zfs / xen failures Thanks again -- This message posted from opensolaris.org
Muhammed Syyid wrote:> removing both folders > /tank/vm/centos54 > /tank/vm > > and rebooting got both the hypervisor up and centos itself running. > Thanks for all your help > > Another question - I was planning on running multiple domU/VMs,> I was planning on using a disk/path command as follows> --disk path=/tank/vm/ubuntu910/disk0,size=70,driver=phy,subdriver=zvol > > Would that work or am I at risk of reverting to the failure mode.Yes, that will work fine.. There are a couple of minor bugs that need to be cleaned up for phy:zvol. One of which you will want to be aware of. The zvol should not exist before you pass that to virt-install. (virt-install doesn''t handle it when it''s already there right now).> Also any insight on why the existing paths (that were created> before the domU disk /zvol was) caused zfs / xen failures No. I have seen failures like that before with zfs. It doesn''t happen that often and I have never been able to characterize what causes them.. MRJ