Comments for Justin's IT and Security Pages

Comment on OpenIndiana and XenServer by Nate Carlson

Nate Carlson — Tue, 09 Apr 2013 13:41:18 +0000

..and pv-grub works great (you can select the boot menu option, change it, etc, just like native hardware); just need to specify the following ‘PV-args’:

(hd0,0,a)/boot/grub/menu.lst

This is on a OpenIndiana install with one virtual disk, and I told it to use the whole disk.

pv-grub is built as described here, with ZFS patches:
https://github.com/aszeszo/pv-grub

Comment on OpenIndiana and XenServer by Nate Carlson

Nate Carlson — Mon, 08 Apr 2013 21:32:31 +0000

..type-o in my comment above; ‘/boot/guests’ should be ‘/boot/guest’, as noted in comments way up above. Just a different error message for me.

In addition, keyboard works after restarting XenCenter. Stupid XenCenter.

I’m also trying to get this working with pv-grub so that updates in the guest will be reflected.. got it to the point where pv-grub boots, but doesn’t find menu.lst; when I paste in the contents of the stock menu.lst from openindiana it boots.

Comment on OpenIndiana and XenServer by Nate Carlson

Nate Carlson — Mon, 08 Apr 2013 21:01:30 +0000

A few notes from my experience on OI 151a7 on XenServer 6.1:

* This works on XenServer 6.1 – however, the kernel and ramdisk must be under /boot/guests, or else you will get an error that says “Internal error: xenopsd internal error: Domain.Could_not_read_file(“”)”. Took me awhile to debug that one, but if you look through xensource.log, you will see messages that the path to /opt/osol is not whitelisted, and see the actual domain construction passing through a blank kernel and bootloader.
* /xpvd/xdf@51712:a worked for me with OpenIndiana 151a7
* Keyboard console doesn’t work for me; I ended up adding the ‘livessh=enable’ kernel parameter while installing, and set up a DHCP server to give the VM a static lease, then SSH’d in as jack to do the install.

Comment on OpenIndiana and XenServer by HellMind

HellMind — Fri, 05 Apr 2013 09:05:19 +0000

pseudo0 at root
pseudo0 is /pseudo
scsi_vhci0 at root
scsi_vhci0 is /scsi_vhci
xdf@51712, xdf0
xdf0 is /xpvd/xdf@51712
/xpvd/xdf@51712 (xdf0) online
xdf@51712: 41943040 blocksNOTICE: zfs_domount: error 19
Cannot mount root on rpool/ROOT/napp-it-0.8h_update_06.28-1 fstype zfs

panic[cpu0]/thread=fffffffffbc60aa0: vfs_mountroot: cannot mount root

Warning – stack not written to the dump buffer
fffffffffbcb5180 genunix:vfs_mountroot+75 ()
fffffffffbcb51b0 genunix:main+136 ()
fffffffffbcb51c0 unix:_locore_start+7e ()

skipping system dump – no dump device configured
rebooting…

Comment on OpenIndiana and XenServer by HellMind

HellMind — Fri, 05 Apr 2013 09:01:28 +0000

unt: error 19
Cannot mount root on rpool/ROOT/napp-it-0.8h_update_06.28-1 fstype zfs

panic[cpu0]/thread=fffffffffbc60aa0: vfs_mountroot: cannot mount root

Warning – stack not written to the dump buffer
fffffffffbcb5180 genunix:vfs_mountroot+75 ()
fffffffffbcb51b0 genunix:main+136 ()
fffffffffbcb51c0 unix:_locore_start+7e ()

skipping system dump – no dump device configured
rebooting…

Comment on OpenIndiana and XenServer by HellMind

HellMind — Fri, 05 Apr 2013 04:28:59 +0000

Isnt working for me, I cant partially boot without all the -B stuff
but I cant do nothing

It ried with all the snapshot i got rpool openindiana and nappit which is my last and and my xcp vm just poweroff alone

Comment on Installing NexentaCP 2 RC1 on XenServer 5 by justin

justin — Fri, 18 Jan 2013 07:14:27 +0000

PV should be a lot faster (overall) than the HVM version. I’m not sure about the hit that you would expect on the disk access, though. It does seem weird that HVM would be faster than the native drivers.

Comment on Installing NexentaCP 2 RC1 on XenServer 5 by Brian

Brian — Fri, 18 Jan 2013 06:35:25 +0000

Whelp, I managed to get this thing up and running. I think the key was that I had to get into that maintenance mode on the built ramdisk, do the syspool import, swap to the boot_archive file for the ramdisk, reboot. I also was missing the zfs-bootfs= (had bootfs=).

One interesting note though:
root@nexpvm:~# dd if=/dev/zero of=/media/test/test.img bs=1M count=1024 conv=fdatasync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 8.5963 seconds, 125 MB/s

root@nexhvm:~# dd if=/dev/zero of=/media/test/test.img bs=1M count=1024 conv=fdatasync
1024+0 records in
1024+0 records out
1073741824 bytes (1.1 GB) copied, 6.36442 seconds, 169 MB/s

On other DD based tests, looks like all the DD tests come back FASTER on the HVM vs the PVM.

Was it your observation that the PVM system was slower or faster than the HVM system?

Comment on Installing NexentaCP 2 RC1 on XenServer 5 by Brian

Brian — Fri, 18 Jan 2013 05:00:30 +0000

The problem above was I had a comma there before boothpath, so it would seem that when i include bootpath=”/xpvd/xdf@51712:a” or any variant of boothpath, I get a kernel panic, enter kmdb, but can’t do anything (no console input). When I exclude bootpath=”/xpvd/xdf@51712:a”, I get the boot CD stuff mentioned above. If I have the CD on the system, then I enter into the installation stuff.

Comment on Installing NexentaCP 2 RC1 on XenServer 5 by Brian

Brian — Fri, 18 Jan 2013 04:41:41 +0000

I updated the PV-args before the -B to include -kv (which gave me verbose output). Two things.

I noticed the first line was:
krtld: Unused kernel arguments: ‘boothpath=”/xpvd/xdf@51712

but continues…

Later on down, I see it online that device by the following lines:
xdf@51712, xdf0
xdf0 is /xpvd/xdf@51712
/xpvd/xdf@51712 (xdf0) online

Appears to mount the ramdisk for xdf@51712:
xdf@51712: 104857600 blocksCD-ROM: Hostname: hardy-core_installcd

Wait a sec… the cd?

mount: Mount point cannot be determined
error mounting /.livecd on
ERROR: svc:/system/filesystem/usr:default failed to mount /dev/fd (see ‘svcs -x’ for details)
… blah …

Not sure if this helps you out… I do have a recorded session of the boot since I can’t copy the console (doesn’t appear that the console=ttya part is working).