[mythtv-users] Is xfs flaky or is it my drive. Fedora Core 2, XFS, Myth .16 and .17

[Rich Hall]

William reportedly babbled:
>
> -----Original Message-----
> From: mythtv-users-bounces at mythtv.org
> [mailto:mythtv-users-bounces at mythtv.org] On Behalf Of tdixon at dixbizserv.com
> Sent: Sunday, March 13, 2005 8:31 PM
> To: Discussion about mythtv
> Subject: [mythtv-users] Is xfs flaky or is it my drive. Fedora Core 2,
> XFS,Myth .16 and .17
>
>
>
> I have a myth box about 6 months old.  Fedora Core 2, new 200GB drive myth
> .16 and .17.
>
> I have the 200GB drive in a volume group for expansion, and its setup XFS
> for /video.  I run ok for a few days, then I'm watching something and it
> stops in the middle, check the logs and I'm getting XFS errors.  Stop myth,
> unmount, xfs_check.. problems get fixed and I run ok for a few days.  Then
> the problem reoccur.
>
> So my questions to the great users of myth..
>
> - Are people using XFS and finding it generally stable or flaky like I find
> it?
> - What are people using as the file system for /video and are you
> happy/unhappy?
> - Do I need to pull this thing apart and do a surface test on the drive?
> - Could there be some things I need to check if you had similiar issues and
> resolved them?
>
> Thank you in advance!
>
> Tom Dixon
> tdixon at dixbizserv.com
>

I hear a lot of people swear that they have solid systems with xfs, and I am not
saying that is not so, but here is some of my experience with it. Mind you this is a
fairly large network cluster (webbot servers) and is growing constantly.. not a
small home net.. your milage may vary.


I have a client that I do a lot of work for who has a growing cluster of (currently
more than 150) FC3 boxes (some early boxes were FC2 and have been moved to FC3)..
AMD 3200-64's with 8 gb of ram per and running 6 drives per box - 1 TB SATA RAID
arrays per box.. The cluster grows by about 6 boxes a week (this is my part). We
started out running ext3 with lvm support as the filesystem because it is dependable
and recovers very well from a crash and all was fine and rock solid.. Someplace
along the line a few months ago one of his minions decided to dump the ext3
filesystem and started using xfs on the newer boxes being added and started changing
the existing boxes over to xfs.. it has been a nightmare.. Lost data, corruption of
the filesystem and a royal pain. The boxes run for a few weeks or more then start to
puke the fs.

After 2 months of trouble we are now migrating the entire site back to ext3. The
slight speed increase over ext3 is not worth the cost of administration for them.
The addition of more memory (from 4 to 8 gb) in each server has brought the overall
performance of the ext3 boxes up to nearly the same performance levels that they
were getting with the xfs filesystems due to the massive caching. In the 8 months I
have worked for these guys, I have replaced about 21 drives out of more than 900
drives currently in use. It is NOT drive failure that has been the issue as the
drives are all RAID. The boxes that so far have been moved back to ext3 are once
again running just fine.

I am just glad I am not the guy who decided that he had to change things as it was
OK before messing with it all. Oh yeah... Aside from the occasional drive failure,
the older ext3 boxes that were not converted are still running without errors. LOL

As I said, your milage may vary.. we just are not getting the stability that others
may see.. but again we are running under some very different conditions.

One of my own primary servers is still running RH 9 ext3 and has an uptime of over
845 days as of this morning.

Good Luck!
-R
--
------------------------------------------------------------------------
 Rich Hall
 rich at netlynx.us
 http://www.netlynx.us/rich/
 ham radio: kf6arx
------------------------------------------------------------------------
 No trees were destroyed in the sending of this message; however,
 a significant number of electrons were terribly inconvenienced.
------------------------------------------------------------------------
 And remember - if it ain't broke, hit it again.