Lustre discuss - Jul 2009 - 1.6.4.2 -> 1.8.0.1 rolling upgrade?

Hi;

We''re interested in doing a rolling upgrade of the Lustre servers for 
one of our production filesystems from 1.6.4.2 to 1.8.0.1.  We have one 
MDS/MGS, and two OSS, no failover.

Rather than upgrade all the server machines at once, we''re considering 
upgrading one machine per day:

* day 1: upgrade MDS/MGS
* day 2: upgrade OSS1
* day 3: upgrade OSS2

 From the 1.8 operations manual (chapter 13), it seems like this should 
be fine to do.  Are there any "gotchas" involved with running a 
combination of 1.6 and 1.8 servers together?

Thanks,
Craig Prescott
UF HPC Center

On Jul 24, 2009  14:10 -0400, Craig Prescott wrote:
> We''re interested in doing a rolling upgrade of the Lustre servers
for
> one of our production filesystems from 1.6.4.2 to 1.8.0.1.  We have one 
> MDS/MGS, and two OSS, no failover.
> 
> Rather than upgrade all the server machines at once, we''re
considering
> upgrading one machine per day:
> 
> * day 1: upgrade MDS/MGS
> * day 2: upgrade OSS1
> * day 3: upgrade OSS2
> 
>  From the 1.8 operations manual (chapter 13), it seems like this should 
> be fine to do.  Are there any "gotchas" involved with running a 
> combination of 1.6 and 1.8 servers together?
In theory, no, but we generally test upgrading the servers together.

Generally there isn''t an admin issue for upgrading servers at one
time, and clients at some other time, so we don''t test every
combination
of MDS/OSS/client.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.