[mythtv-users] mythbackend.log is buffered?

Jerome Yuzyk jerome at supernet.ab.ca
Mon Oct 26 19:58:49 UTC 2015


On Monday, October 26, 2015 03:03:19 PM Eric Sharkey wrote:
> On Mon, Oct 26, 2015 at 1:04 PM, Jerome Yuzyk <jerome at supernet.ab.ca> wrote:
> > This is what I've been given for Myth from the RPMFusion packages. I think the 'nocreate' option is what is not creating a new log file, unlike every other logfile on any of my Fedora systems. Not sure why it's there, except maybe it's expected that the restarting mythbackend will create it. Is that the logic? Is there a reason I shouldn't remove that option?
> 
> I would advise against removing that option.  Removing it won't help
> with the problem you reported.
> 
> 
> >     firstaction
> >         /bin/systemctl kill mythbackend.service --signal=HUP --kill-who=main 2>/dev/null || true
> >     endscript
> 
> That is bizarre.  I can't think of any reason to send mythbackend a
> HUP signal in the firstaction section (before logs are rotated).
> 
> If you send a HUP signal, and it reopens the existing log, and then
> it's rotated, it will end up writing to the old log file.  That's not
> at all what you want.  You need to send the HUP signal after rotation
> is done.

That's what RPMFusion packages gave me and I never touched it, so maybe that dev can comment. 

Logrotate has always Just Worked and I've tail-ed a lot of rotated logs and just stop/restart to continue tail-ing. Tail has never interfered with logging. Even on my old 0.21 box. I've never had to dive into how it works. Somewhere along the way that changed for Myth - it creates files like mythbackend.20151026194647.23812.log, and creates a new log file at every restart.

I just restarted mythbackend and as Bill predicted I have a logfile again.

-- 
A little of Jerome's MythTV World: http://mythtv.bss.ab.ca


More information about the mythtv-users mailing list