[Lustre-discuss] More detail regarding soft lockup error

[Andreas Dilger]

On 2010-08-19, at 10:49, Brian J. Murrell wrote:
> On Thu, 2010-08-19 at 10:09 -0600, Andreas Dilger wrote: 
>> 
>> If you increase the size of the MDT (via resize2fs) it will increase the number of inodes as well.
> 
> Andreas: what is [y]our confidence level with resize2fs and our MDT?
> Given that I don't think we regularly (if at all) test this in our QA
> cycles (although I wish we would) I personally would be a lot more
> comfortable with a backup first.  What are your thoughts?  Unnecessary?

Always have a backup of the MDS, even if you are NOT doing an inherently risky process like potentially rewriting all of the metadata in the filesystem...  I keep two full "dd" copies of my MDS, alternating days, given that the space required is so small.

Even if there is a large MDS with short-stroked SAS drives or SSDs in RAID-1+0, keeping a handful of slow 1.5TB SATA drives attached just for backups makes a lot of sense, and costs a few hundred dollars.  They don't need to be dual-ported (or even more than RAID-0 or LVM concatenated volumes).  Because the "dd" backup and restore is entirely linear IO the SATA drives will give very good performance.

That said, any ext* filesystem formatted in the past 5 years can normally do a resize of 1024x without actually having to scan/rewrite the filesystem metadata.  I can't say that I've had to do any MDT resizing, but I've resized my OSTs a bunch of times w/o ill effects.  That isn't to say that Oracle tests this, just my personal observation from my home setup.

Cheers, Andreas