Re: [HPDD-discuss] kernel panic when upgrading from lustre-zfs 2.3.49 -> lustre-zfs 2.4
Nate,
It looks to me like you were running a very experimental branch Lustre.
I fear that your best bet is to create a new Lustre filesystem from
scratch starting at a non-experimental version of Lustre. Lustre made a
few incompatible on-disk format changes in the lead up to the official
Lustre 2.4 release. Even if the crash you are hitting is not one of the
known problems, those other known issues are waiting to bite you next.
Chris
On 09/06/2013 01:09 AM, Nathaniel Everett Garver-Daniels wrote:
Hello,
Last year around this time I set up a lustre / zfs system based on
zfs 0.6.0_rc3_31chaos
lustre 2.3.49.92
that I built from source. It's been chugging along merrily with a
combined MGS/MDS and Three OSSs, each with one OST. For some other
reasons I had to upgrade the other machines in my cluster from Redhat
6.3 to Redhat 6.4 and it seemed like a good time to try and update to a
more official release. I followed the instructions on zfslinux.org for
adding their repository and installed zfs 0.6.2 and lustre-2.4.0.
After un-mounting all of the clients / OSTs and the MGS/MDS I restarted
the system with the new lustre and zfs installed. The zfs filesystems
looked OK, but when I tried to "mount -a -t lustre" it had a kernel
panic. I haven't been able to get a dump off of the machine yet, but I
was wondering if anyone else has run in to this type of problem when
doing an upgrade with zfs osds.
Also, any recommendations on trying to troubleshoot this while I leave
the current system running would be appreciated. My plan is to send a
zfs snapshot of the MGS/MDS to another machine and try upgrading it
there. I'm wondering if I might need to do a "zpool update" before I
try and mount the lustre filesystem.
Thanks,
-Nate Garver-Daniels
Systems Administrator
_______________________________________________
HPDD-discuss mailing list
HPDD-discuss(a)lists.01.org
https://lists.01.org/mailman/listinfo/hpdd-discuss