[mythtv-users] processes getting wedged in D state on FC6 kernel 2944?

[Kevin Kuphal]

Robert Macaulay wrote:
> While running updated FC6 x86_64 system with myth 0.20.1-155.fc6.at
> packages from atrpms, I'm beginning to see backend processes getting
> stuck in the un-interruptible waiting on IO state(The dreaded "D"
> state).
>
> It is running  2.6.20-1.2944.fc6. Tried backrevving to 2933. No difference.
>
> There are no errors in dmesg prior to the death. Things like "sync"
> will block forever. Attempting to stop myth hangs. Load goes high(11
> last time), nothing on CPU though(all cpu time is idle - no io wait,
> user, or system).
>
> Basically, something is wedging access to disks. About 10 minutes
> before the wedge, I got this in the backend logs
> 2007-05-03 20:34:50.607 TFW, Error: Write() -- IOBOUND begin cnt(9400)
> free(1151)
> 2007-05-03 20:34:52.650 TFW, Error: Write() -- IOBOUND begin cnt(2048)
> free(2047)
> 2007-05-03 20:44:52.752 Expiring Crossing Jordan "Dead Again" from Wed
> May 2 20:00:00 2007, 95 MBytes, forced expire (LiveTV recording)
> 2007-05-03 20:44:52.825 autoexpire: Expiring Program: Expiring
> Crossing Jordan "Dead Again" from Wed May 2 20:00:00 2007, 95 MBytes,
> forced expire (LiveTV recording)
> 2007-05-03 20:45:26.100 MainServer::HandleAnnounce Monitor
> 2007-05-03 20:45:26.110 adding: shiner as a client (events: 0)
> ==force reboot here ==
> 2007-05-03 20:56:13.187 Using runtime prefix = /usr
>
> It was when the wife sat down to watch on a remote frontend. Its
> running on a LVM JFS backend. Its definitely kernel. Looking to see if
> anyone else has seen this.
>   
I have the same thing after doing quite a few changes to my system but 
eventually getting to 2.6.20.4.  Any time I get an IOBOUND mythbackend 
seizes and if I try to stop it goes defunct.  I have pretty much 
eliminated the IOBOUND by moving my /var partition with logging and 
mysql to a separate drive from my LVM but I also am using LVM and JFS.  
Can't say that my situation helps you but my system is stable when 
IOBOUND is eliminated

I have 2 HD and 2 SD tuners and run one commflag job at a time, FYI

Kevin