[mythtv-users] Mythbackend taking 50% CPU w/PVR150 or 250 when recording

[Michael T. Dean]

On 02/27/2006 03:56 PM, Stuart Larson wrote:
>> On 02/27/2006 11:30 AM, Stuart Larson wrote:
>>     
>>>   First off, system info:  Myth from 0.19-fixes branch, checked out from
>>> 2/24/06.  Running the backend on FC4, IVTV 4.2, and two PVR cards - one
>>> is a 150, the other a 250.  My storage dir is on a 4-disk LVM, formatted
>>> in ReiserFS.  PC is an AMD Sempron 3100, with 1GB of ram.  DMA is
>>> enabled on all disks.
>>>    I've had no problems with recording up until this weekend.  What I
>>> started seeing is when mythbackend is recording, it starts taking up
>>> around 50% CPU.  Normally I wouldn't think this would affect playback
>>> on the local frontend (they are on the same machine), but if either of
>>> the cards were recording, I'd start seeing skips and pauses,
>>> corresponding to mythbackend's CPU usage.  The actual recordings, onced
>>> finished, play back fine, it just appears that playback of other media
>>> (the entire box was slow) suffered because of mythbackend taking up so
>>> much CPU.  Before this, the highest I saw mythbackend take was around
>>> 10%.  If it's relevant, I'm recording at 480x480, around 3500 bitrate,
>>> and I have 1 GB reserved for files.
>>>   What I noticed (and from searching the lists, it looks like this was
>>> an
>>> issue brought up but not really resolved), was that this started
>>> happening when my storage space went under 10G available.  Once I
>>> cleaned up some old recordings and got more space available, mythbackend
>>> went back to using 3-10% CPU when recording...  I did check the
>>> mythbackend log, and I did not see anything out of the ordinary - no
>>> messages about expiring old programs.  I ran it on --verbose all (whoah
>>> lots of info!) and also did not see anything out of the ordinary.
>>>   Has anyone seen this issue before?  I was able to resolve this just by
>>> cleaning out old recordings, but I wonder if there's a setting I'm
>>> overlooking that would cause/fix this type of behavior.
>>>
>>>       
>> IIRC, XFS is known to behave poorly with a "nearly" full filesystem.
>> What filesystem are you using?
>>     
> Sorry, it was a badly formatted email.  I'm running ReiserFS on the LVM.
>   

No.  I'm sorry.  My fault.  I totally missed that...  Sorry for the noise.

Mike