[mythtv-users] Tracing Instability

[blind_Pete]

On Mon, 15 Aug 2011, Ken Smith wrote:
> Brian J. Murrell wrote:
> > On 08/13/2011 05:32 PM, Ken Smith wrote:
> >    
> >> Hi, I'm on the trail of some instability (random hanging) of my
> >> Mythbackend.
> >>      
> > Welcome to the club.
> >    
> {snip}
> >    
> >> I've set up a script based on
> >> http://tweakingmythtv.wordpress.com/2011/01/11/mythbackend-deadlocks-and-auto-restart-script/
> >> to catch the problem and restart the backend.
> >>      
> > Interesting.  The only problem with that script is that it might restart
> > the backend while it's recording something (mine seems to only lock up
> > when recordings are in progress) and no clients are needing to connect,
> > needlessly interrupting a recording.
> >
> >    
> Yes it does that. I had an hour long recording in 6 parts the other 
> day... A few days ago I ran a backup app that loaded the machine very 
> heavily and the Mythbackend was restarted, by my heartbeat script, on 
> average every 4 mins over a period of 4 hours. Fortunately nothing was 
> scheduled to record at the time. In the end I stopped myth and let the 
> backup complete. It was seeing that happen that pointed me toward the 
> theory that the mythbackend hangs might be related to CPU load. It's on 
> 14,000 bogomips at the moment.

I'm running boinc, so my machine is always heavily loaded, that might be 
contributing to my problems.  

> >> mythtv-backend-0.24.1-275  (atrpm's build)
> >> mysql-5.0.77-4.el5_6.6
> >> 2.6.18-238.12.1.el5.centos.plus 64 bit
> >>      
> > So, now we have one on Ubuntu (me) one on Gentoo (IIRC) (warpme) and one
> > on CentOS.  It is an interesting data point that it's not isolated to a
> > single distribution.

Add Mandriva 2010.2 with plf backports to your list.  

[root at live ~]# rpm -q mysql
mysql-5.1.58-0.1mdv2010.2
[root at live ~]# rpm -q mythtv-backend
mythtv-backend-0.24-20110303.1plf2010.2
[root at live ~]# cat /etc/release 
Mandriva Linux release 2010.2 (Official) for x86_64

> >> Any suggestions as to what I should look at to narrow this down.
> >>      
> > Well, you can start by searching the very recent archives for a long
> > thread about this issue.

Do you mean, "MySQL related BE deadlocks - collective wisdom needed"?  

> 
> Yes, re-read those and the bug reports today.
> 
> As a datapoint, I am using the OTA program guide and I don't run 
> commflagging. I also have a few applications that make occasional use of 
> other databases in mysql.
> 
> As another datapoint, mysql's files are on a raid5 ext3 file system 
> hosted on a 3ware controller with IDE physical disks. The video files 
> are on a single non raid sata disk. The performance of the 3ware isn't 
> that spectacular and I see average 8% iowait according to sar. Also, I 
> have noticed that the machine has used the swap file in the last few 
> days so I have another Gig of ram on its way.
> 
> Currently I've set my mythbackend and mysql to run at nice -3 to give 
> those processes priority over the slew of other things that server does. 
> No restarts since I did that about 12 hours ago.
> 
> How different is the 0.23 schema from 0.24? All I would need to save is 
> my recorded programs. The rest can be recreated. Is a downgrade at all 
> possible until maybe this problem is fixed in 0.25?
> 
> For now I'll see if my nice -3 setting and another gig of ram helps 
> before I do anything more drastic.
> 
> Thanks.... Ken