[mythtv-users] mythlogserver multiple copies high cpu

[Craig Treleaven]

At 8:03 PM +1030 3/7/13, David Farmer wrote:
>On Mon, Mar 4, 2013 at 1:23 PM, David Farmer 
><<mailto:remrafevad at gmail.com>remrafevad at gmail.com> wrote:
>On 02/03/13 17:13, David Farmer wrote:
>>I have been having this as an ongoing problem since I installed 
>>MythTV 0.26 on an Arch Linux x86_64 box. There is a 
>><http://code.mythtv.org/trac/ticket/11230>current bug report 
>>(ticket #11230) but it's priority is marked at low, I'm hoping that 
>>this could possibly change! Trouble is, I don't know how to 
>>reproduce it. As the subject says, I get multiple copies of 
>>mythlogserver and they smash the cpu at a combined total of almost 
>>100% for each core.
>>
>My typical MythTV usage is to have it wake itself, record a program 
>and then use a HandBrakeCLI user job to transcode this, and finally 
>shut itself down. Typically a 2 hour SD recording takes 2 hours to 
>transcode using my settings, and a 2 hour HD takes about 6 hours. 
>But since 0.26 I get up in the morning and the the user jobs aren't 
>finished and there are around 4 copies of mythlogserver all running, 
>using so much cpu that HandBrakeCLI only used 0.3%. But the worst 
>example was a few days ago, I had 14 instances all at the same time.
>
>Adding "system" to the verbose settings should cause a log entry each
>time mythlogserver is started up. A line like: 
>
>>2013-01-22 19:31:10.371307 I [401/28931] Logger system-unix.cpp:868 
>>(Fork) - Managed child (PID: 402) has started! * 
>>command=/opt/local/bin/mythlogserver --daemon --verbose none,system 
>>--logpath /opt/local/var/log/mythtv.26 --loglevel debug, timeout=0 
>
>I've changed the mythlogserver parameters as suggested, but as yet 
>haven't had the problem appear. Will these new settings survive a 
>reboot, or do I need to specify them each time?
>
>Ok, so it took a while but I have a recurrence of the problem. A 
>quick description of the events before and after I noticed it.
>
>1. I logged into the HTPC via ssh to see how many mythlogserver 
>processes were running, only 1 and cpu usage was less than 1%. I 
>then logged out.
>2. I opened mythweb from the same pc as I did the ssh. I cancelled a 
>few upcoming recordings that were scheduled via a custom search 
>rule. Closed Mythweb.
>3. Ate dinner
>4. I sat in front of the HTPC and checked how many mythlogservers 
>were open, 7 or 8 (forgot to count!) of them and cpu usage was 
>maximum.
>
>The log file is huge though, because all this time Shepherd (my 
>Australian program listings grabber) was running as a cron job. An 
>immediate issues I've noted in the log is
>
>"Mar 07 18:51:20 david-htpc systemd-coredump[2706]: Process 2672 
>(mythpreviewgen) dumped core."
>
>Not sure which other messages are important, how should the log be uploaded?

If you grep the logs, (say for "bin/mythlogserver"), you ought to 
find several instances of mythlogserver being started while you were 
at dinner.  Maybe they are all in your mythbackend log, but they 
could also be in the logs of any other program that ran in that 
period:  mythcommflag (?), mythtranscode, mythpreviewgen, etc.

Craig