[mythtv-commits] Ticket #8199: backend lockup on recording expiry

[MythTV]

#8199: backend lockup on recording expiry
-------------------------------------+--------------------------------------
 Reporter:  michael_zanetti@…        |        Owner:  danielk   
     Type:  defect                   |       Status:  new       
 Priority:  trivial                  |    Milestone:  unknown   
Component:  MythTV - Recording       |      Version:  0.22-fixes
 Severity:  medium                   |   Resolution:            
  Mlocked:  0                        |  
-------------------------------------+--------------------------------------
Changes (by michael_zanetti@…):

  * status:  closed => new
  * resolution:  invalid =>


Comment:

 Replying to [comment:10 danielk]:
 > The i/o waits indicate an OS problem. ext2/3 are known to have serious
 problems handling large files. And ext4 is known to have serious problems
 as a system disk. Absent further evidence I'm assuming you are using one
 of these filesystems in an application for which it is not suited and am
 closing as invalid.

 There are no i/o waits... It happens also when the system is idle.

 Nevertheless, I have changed my recording partitions filesystems to XFS as
 suggested by the wiki. The problem still persists.

 Btw. I have upgraded to 0.23-fixes in the meantime... Still the same.

 Today, when it locked up again I had a look at the logs and hope I have a
 more detailed description now. Please see the attached log "mythbackend-
 lockup.log" in combination with this comments:

 - The backend should have started to record a show starting at 15:10. I
 have set it up to start 3 minutes earlier for every show plus 3 minutes
 for this specific one. So the actual start should have been at 15:04. The
 machine woke up on RTC alarm and freshly bootet up from poweroff. At
 15:02:02.493 the backend came alive for the first time.

 - The recording was successfully started. However, it is only 4 seconds
 long!!

 - Then, at 15:18:36.054 the backend seems to have restarted itself. I
 haven't been at home so no one could have restarted the backend. Is this
 possible? I haven't set up a watchdog to do such tasks.

 - At 15:18:38.840, there is again this message. "~MythContext waiting for
 threads to exit." After that there isn't much activity from the backend
 any more. There are still a few messages generated to prove that the
 process is still alive.

 - The (now locked up) backend missed a recording at 18:00 and another one
 at 18:25 without any notice in the logs.

 - About 19:30 I turned on TV, started the frontend out of mythwelcome and
 got to the system information screen. The first page, the one about EPG
 data, has been shown successfully. When I pressed "down" to go to the
 schedules-page also the frontend locked up. I first killed the frontend,
 then restarted also mythwelcome. This produced the few last log entries
 starting at 19:33.

 Then, at 19:35 I restarted the backend. Its running again fine since then.

-- 
Ticket URL: <http://svn.mythtv.org/trac/ticket/8199#comment:11>
MythTV <http://www.mythtv.org/>
MythTV