[mythtv-users] Defunct Mythbackend

[Kevin Kuphal]

Matt Rude wrote:
> Kevin Kuphal wrote:
>> Matt Rude wrote:
>>  
>>> Kevin Kuphal wrote:
>>>    
>>>> Matt Rude wrote:
>>>>  
>>>>      
>>>>> I have be having this problem fro the last few weeks where after a 
>>>>> day or so my Mythbackend demon will go defunct and not let me 
>>>>> restart the demon
>>>>>
>>>>> /sbin/service mythbackend restart
>>>>> fails every time. the system wont even let me reboot when I try 
>>>>> the system will just sit there like as if I didn't tell it to reboot.
>>>>> ps -eaf shows me
>>>>> root      3673     1  5 00:41 ?        00:41:27 [mythbackend] 
>>>>> <defunct>
>>>>>
>>>>> anyone have any idea what I can look at.
>>>>>
>>>>>
>>>>> This is a Fedora Core 6 system with an Nvida 7300
>>>>>             
>>>> I have the same problem with a newly updated kernel system.  It 
>>>> appears to happen after an IOBOUND message from the backend.   What 
>>>> is your disk layout?  Is your /var folder on the same 
>>>> filesystem/drive that your recordings are occuring on?  How many 
>>>> and what kind of tuners are you using?  I was able to prevent this 
>>>> by moving my /var partition (mysql and logging) to a separate 
>>>> drive.  Now I never have this problem.
>>>>
>>>> Kevin
>>>>         
>>> My recording directory's (/mythtv/Recorded & /mythtv/Recorded2) are 
>>> on different drives
>>> but the rest of the system is on a 3rd 80 gig system drive.
>>>
>>> I am using a PVR-350 (tuner only no TV-out) and a PVR-500 for the 
>>> tuners
>>> TV-Out via a Nvidia 7300
>>>
>>> so do you currently have the var directory and mysql running on 
>>> separate drives or did you remove your recordings from that drive?
>>>     
>> In my situation, part of my LVM (JFS) array was built from my boot 
>> drive.  I added a separate drive and mounted it as /var.  Once I did 
>> that, all my problems went away.   In my case though, I was recording 
>> 2 HD streams and 2 SD streams plus commercial flagging.   It doesn't 
>> sound like your situation is exactly analogous to mine.  Do you get 
>> IOBOUND messages in your backend logs?
>>
>> Kevin
>>
>>   
> Yes
>
> grep IOBOUND mythbackend.log
> 2007-05-15 13:52:24.965 TFW, Error: Write() -- IOBOUND begin cnt(2048) 
> free(2047)
> 2007-05-15 13:52:25.334 TFW, Error: Write() -- IOBOUND begin cnt(2048) 
> free(2047)
> 2007-05-15 13:52:25.513 TFW, Error: Write() -- IOBOUND end
> 2007-05-15 13:52:25.525 TFW, Error: Write() -- IOBOUND end
I think you will find that the defunct status corresponds to those 
messages.  If so, you need to find a way to eliminate them.  You could 
try things like lowering the number of jobs that can run at the same 
time, make sure your hard drives are optimized (DMA on in Linux, 80-pin 
IDE if they are IDE drives, etc).

Kevin