[mythtv-users] mythbackend periodically locks up at 100% CPU since upgrading to latest 0.27.4 fixes

Paul Netherwood pnetherwood at email.com
Sun Nov 23 17:54:34 UTC 2014


> Ever since updating to the latest fixes I've been having trouble with the backend locking up at 100% CPU. The front ends then become unresponsive and nothing can connect to the backend including Mythweb. After 5-10 minutes it frees up again. It tends to be worse when recordings are going on. In the logs all I can see are errors to do with MythSocket and WriteStringList although I think is is more to do with sockets timing out because the backend is unresponsive.
> 
> Digging a bit deeper I used gcore and gdb to have a look at what was going on and I noticed lots of Scheduler threads locked on QMutex. I've managed to reduce the problem somewhat by doing the following:
> - Disabling EIT scanning (I use xmltv anyway) because the stack trace was showing EIT threads locked on QMutex
> - Removing multirec from my capture cards which reduces the number of tuners and therefore the number of scheduler threads
> 
> It still locks up but the duration is shorter. I'm think a bug has been introduced in the latest fixes possible since 0.27.3. The problem first started occurring when I did an update around the time when the bash Shellshock bug was announced. Everything was fine before then.
> 
> I also noticed some of my HD recordings are now being cut short due to maximum buffer size exceeded in DBVRead. It says that it "indicates your disk performance is insufficient to deal with the number of on-going recordings, or you have a disk failure" but I suspect its down to the backend locking up since it never used to happen.
> 
> I have logged a new ticket which has more detailed information (see https://code.mythtv.org/trac/ticket/12314). I was wondering if anyone else had had the trouble or had found a workaround.
> 
> Paul
> 
> ------------------------
> Backend system details:
> Software:
> Ubuntu 14.04 (3.13.0-37)
> Mythtv 0.27.4
> Mysql 5.5.40
> Hardware:
> Intel Core i3
> 2gb RAM
> 4 x Nova-T
> 1 x TBS 6920
> 4 x 3TB harddisks
> 245GB SSD (system disk)
> 
> Frontend system details:
> AcerRevo running same version of Ubuntu and Mythtv as backend but from the Mythbuntu distro.


I think this problem has something to do with multirec. I have 5 cards, 4 DVB-T cards each with multirec set to 2 and a DVB-S2 card with multirec set to 5. I had no problem with this configuration until I upgraded to 0.27.4. The time locked out was reduced when I reduced the multirec on the DVB-S2 card from 5 to 2. Once I removed all of the multirec and set every card to 1 the problem does not occur. This effectively reduces the number of encoders from 13 to 5.

I wouldn't say the problem is resolved but reduced. I noticed some changes to the code to do with multirec in scheduler.cpp. 

Has anyone else had this problem?


More information about the mythtv-users mailing list