Doomshammer's Weblog

As I am still running a very old version of Solaris 10 on one of my boxes, I wanted to verify if an upgrade to one of the current builds works w/o issues. So I downloaded Solaris Nevada build 64 and burned on DVD.

As expected the installer directly noticed that there is already Solaris installed on my disk, so it asked if I wanna upgrade. I selected "Upgrade" and the installer started working. It took about 2.5 hours, but no problem occured. The system rebooted and finally snv b64 started booting. I was very surprised that even the RAID-1 on the two root disks was still intact and that the system bootet from it - awesome

After logging in, I noticed that the zpool (where my home directory lifes) was broken. All drives were unavailable and so I wasn't able to take it online again - probably as the ZFS pool was legacy version 3 and the current version of b64 is v6. Anyhow, as ZFS is pretty smart it was very easy to recover it...



I booted into single user mode (-m milestone=single) and removed the ZFS cache file (/etc/zfs/zfs.cache). Then I continued the boot sequence by running svcadm milestone all, to get into the multi-user milestone. A 'zpool status' confirmed that there was no zpool available anymore. Now I simply executed 'zpool import pool1' and my pool was online and healthy again - followed by 'zpool upgrade pool1' my pool was upgraded to v6 and that's it!



Love it! Now i'm safe to upgrade my old system, as my test-system upgrade worked better than expected

• Comments (0)
• Trackbacks (0)

Some days ago I downloaded the latest Solaris Express Community Release (build 70) as I wanted to try out booting from a ZFS root. But after the installation was done, I ran into some troubles.

The installer automatically rebooted after the installation was done, GRUB came up as expected, but when I selected the corresponding SXCR b70 kernel, I ended up in an error I haven't seen before.



I asked in #opensolaris on Freenode, but nobody was able to provide help. So I tried figuring out what's wrong. I checked bootenv.rc if bootpath was set correctly, I added root(hd0,0,a) to the menu.lst of GRUB, reinstalled GRUB and mboot to the MBR but all I did, didn't help. Finally after re-reading the error message several times, I finally noticed the "reading beyond end of ramdisk". Sounded somehow odd to me - why should the bootload read beyond the ramdisk?

So I booted into failsafe again, mounted c1t0d0s0 to /a and started rebuilding the ramdisk:

Followed by init 6 which directly updated boot_archive (which is good, as this means that the system noticed, that the ramdisk with the boot image has changed).

Indeed the rebuild of the ramdisk helped, as - after the box has been rebooted - the system came up as expected.

• Comments (0)
• Trackbacks (0)