[mythtv-users] IOBOUND errors after kernel upgrade from 2.6.31 to 2.6.32
beww at beww.org
Fri May 7 16:29:22 UTC 2010
On Friday 07 May 2010 10:16:22 am Tom Dexter wrote:
> On Sat, May 1, 2010 at 2:17 PM, Tom Dexter <digitalaudiorock at gmail.com>
> > This is interesting...I'm finding a lot of stuff about a change to
> > EXT4 in 2.6.32 that does in fact introduce a performance regression:
> > http://www.phoronix.com/scan.php?page=article&item=linux_perf_regressions
> > Interesting. If I start getting the IOBOUND errors again once the
> > drive is more full I may try adding that nobarrier mount option and
> > see if they go away. My EXT4 partition is only used for recordings
> > (nothing where data integrity is critical like the DB) where loosing a
> > cached write in a power outage is no issue...hell...the whole
> > recording is worthless at that point.
> > Tom
> After a lot of testing here's what I've ultimately discovered: The
> kernel changes that are causing me problems are the changes that
> occurred between linux-2.6.31-gentoo-r6 and
> linux-2.6.31-gentoo-r10...so this wasn't specifically with 2.6.32.
> When I tried linux-2.6.31-gentoo-r10 I actually got some IOBOUND
> errors when recording only one HD show with nothing else going on even
> with good ext4 options. For the last two days I've been running
> 2.6.31-gentoo-r6 again, and even with three shows recording, the
> frontend watching another, and two of the recordings commflagging in
> real time it worked like a clock both nights.
> The ext4 changes referred to in the above link were made between those
> two versions. The specific issue discussed there however should be
> resolved with nobarrier, which didn't resolve it for me...nor did
> changes such as data=writeback,commit=60 etc. As a matter of fact, it
> looks to me as though the changes to the ext4 code between those
> versions was way more extensive than the changes going to 2.6.32.
> They have to have been significant seeing how linux-2.6.31-gentoo-r6
> works for me even using the default ext4 settings, which is basically
> just asking for trouble.
> If it weren't such an awful undertaking I'd try switching back to
> ext3...but who knows, they may have changed that in ways that would
> cause me problems as well. At this point I guess I have to stick with
> linux-2.6.31-gentoo-r6 unless I can figure out a good way to shoe-horn
> a third hard drive into my Dell 4600 for the OS partition. I'm
> assuming that's most likely at the root of all this...have the
> os/database partition sharing a drive with my recording LVM...I'd sure
> hate to go through all that, risking heat/power issues etc, and have
> *that* not fix it.
From what I've read, the kernel folks seem determined to push ext4 out the
door, ready or not, I'm not sure why.
Making it the default filesystem was ill-conceived IMHO. Granted it will get
more eyes looking at it, and thus get more bug reports, but the negative
publicity of problems like you mention will cause many people (like me) to
stay away from ext4, especially for the OS and its root filesystem, that's just
too important to be used as an experiment.
I'd guess going to ext3 would solve your problems, but I can't promise that of
As for "shoehorning", you might use a CF card with a CF-to-SATA adapter just
to test, those things can be "shoehorned" in just about anywhere.
More information about the mythtv-users