[mythtv-users] IOBOUND errors after kernel upgrade from 2.6.31 to 2.6.32

[Tom Dexter]

On Tue, Apr 27, 2010 at 4:00 PM, Yan-Fa Li <yanfali at gmail.com> wrote:
> CFQ is now the default I/O scheduler and it favors interactive rather
> than server loads by default.
>
> http://lwn.net/Articles/355987/
>
> On Tue, Apr 27, 2010 at 7:16 AM, Tom Dexter <digitalaudiorock at gmail.com> wrote:
>> On Tue, Apr 27, 2010 at 10:13 AM, Tom Dexter <digitalaudiorock at gmail.com> wrote:
>>> On Sunday I did various updates on my Gentoo frontend/backend systems.
>>>  Once of them was an upgrade from the 2.6.31 kernel to
>>> 2.6.32...specifically sys-kernel/gentoo-sources-2.6.31-r6 to
>>> sys-kernel/gentoo-sources-2.6.32-r7.
>>>
>>> I noticed a few skips in some recorded programs last night, like there
>>> were small pieces missing.  I discovered that over the two nights
>>> since that upgrade I've gotten several IOBOUND errors while recording.
>>>  If I count just the "IOBOUND begin" messages I see 19 of them.
>>>
>>> I just unzipped all my old logs and can see that prior to that (all
>>> the way back to February) I've had non at all.  There wasn't much of
>>> anything else involved in the upgrade that I would expect to cause
>>> those.  I've rebooted the backend to the 2.6.31 kernel and will see
>>> how it goes tonight (I have a lot of recordings scheduled).  That
>>> should tell me for sure whether or not the kernel is the issue.
>>>
>>> Is anyone aware of any new kernel settings that might affect
>>> this...scheduler changes etc??
>>>
>>> Thanks.
>>> Tom
>>>
>>
>> I forgot to mention...I'm using EXT4 on my recording drive.  I
>> converted from EXT3 at the beginning of March and that's been working
>> fine.  Also, that's on an LVM2 volume.
>>
>> Tom
>> _______________________________________________
>> mythtv-users mailing list
>> mythtv-users at mythtv.org
>> http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
>>

The CFQ sceduler is enabled:

CONFIG_IOSCHED_CFQ=y

...but it's not the default:
CONFIG_DEFAULT_AS=y

...which is exactly as it is in 2.6.31...as a matter of fact, all
kernel parameters with SCHED or DEFAULT anywhere in the parameter name
are the same in both.

By the way, my recordings tonight (three HD shows at a time with one
commflagging) are going without a hitch tonight with the 2.6.31
kernel.

I wonder if anything changed in EXT4??

Tom