[mythtv] Jobqueue problem -- commercial flagging running on wrong host

[David Shay]

I am having an issue with the jobqueue that I don't know if anyone else has
seen.  I am using CVS as of last night, but have had the problem for 2-3
weeks, probably.  In many cases, the system decides to run commercial
flagging on the frontend where it was not recorded.  My master
backend/frontend is a EPIA M10000 and I have a slave backend P4/3Ghz.  These
do not share an NFS file share -- they each have independent storage.  They
are also connected via a wireless network, which is normally not a problem,
but recently it seems that shows recorded on the master backend are getting
commercial flagged by the slave backend, all using the MythTV file transfer
protocol.  I know that that was a recent addition to be able to do that, and
if I wasn't on a wireless network it might work really well, but..

On both the master (epiamythbox) and slave (desktop) backends, in setup, I
have the flag set indicating "run jobs only on original recording host" with
1 maximum simultaneous job, with low CPU usage.  I am only allowing for the
commercial flagging job and no user jobs.

For example, this morning, a program finished recording at about 9:35 AM.
The slave backend "desktop" in it's log (turned on all/verbose logging), I
see:

9:35.27 jobqueue:getjobsinqueue found 'flag commercials' job in queued state
9:35.27 jobqueue: currently running 0 jobs
9:35:27 jobqueue: processing 'flag commercials'  job for xxx , current
status is queued
9:35:28  ANN Playback desktop
9:35:28 ANN Filetransfer  desktop[]:.....
9:35:28 Started commercial flagging for xxx
9:35:28 commflag: Commercial flagging started
9:35:29  QUERY_FILETRANSFER 20[]:.....
etc.

At the same time in the master backend log (epiamythbox)  I see:
9:35:27 MainServer::HandleAnnounce playback
9:35:27 adding: desktop as a client
9:35:28 Mainserver::HandleAnnounce filetransfer
9:35:28 adding: desktop as a remote file transfer
9:35:28 31  OK[]:[]20[]...


Don't know if it's relevant, but I had the jobqueue interval set to 300
seconds on the master backend and 60 seconds on the slave.  I have since set
that to 300 seconds on both.  Don't think that should control where it runs,
though.