OT: BEWARE filesystem changes!

[Brian K. White]

On 3/29/2014 10:31 AM, Fairlight wrote:
> Just a general heads-up.  If anyone is storing -large- amounts of files on
> an older system using reiserfs, be -very- careful when migrating to ext4.
>
> Reiserfs doesn't actually have a hard inode limit, as far as I can tell.
> However, ext4 does.  This becomes problematic when someone has stored over
> 6.5 million inodes on the old system, you go to do an rsync, and run out of
> "disk space" at only 57% of used -space-.  However, you've entirely
> exhausted the inode table for the filesystem.
>
> Even more fun, reiserfs not only doesn't have limits, it doesn't bother to
> keep track of how many inodes there are.  If you weren't using LVM on the
> originating system, it's a complete pain to try and find out how many are
> in use.
>
> There are ways to tweak the ext4 filesystem to increase the inode ratio at
> creation, but you cannot simply allocate/add more inodes later.  I'm in the
> process of finding out if that number grows if you resize the partition, or
> barring that, add a new volume via lvm.  No idea yet.  I'd call the whole
> thing a bad design decision, but I didn't write ext4.
>
> Just be aware of the issue.  I've never personally seen it happen before
> today, but I'm not the only one to be bitten by it.
>
> Common wisdom says avoid more than 500 files in a directory, for
> performance's sake.  Unfortunately, a directory heirarchy which breaks
> things down by year, month, day, hour, minute uses a -hell- of a lot of
> inodes before it even gets to the files underneath.  (Think Postfix style
> heirarchy, but applied to files being stored and referenced by filePro
> records, which is exactly what someone did, and how it's relevant.)
>
> Just a heads-up.
>
> mark->
>

Saving this for next time someone asks me why I still use reiserfs...

-- 
bkw