[mythtv-users] Recordings marked with "deletepending"

[Michael T. Dean]

On 08/03/2014 09:43 AM, Greg Oliver wrote:
> I rarely use mythweb, but this morning I was looking for a video I 
> archived so I fired it up.  One thing that caught my eye was that my 
> recordings showed a series that I knew I had deleted long, long ago.  
> This (and a couple other) series do not show up in the frontends.
>
> Upon investigation, it was due to the "deletepending" column being set 
> in the recorded table.

These should never, ever show up in MythWeb (or mythfrontend).  If they 
are, there's a bug in MythWeb.  The only notice you should ever have of 
them is some amount of "Space used by deleted recordings" being reported 
all the time, even after you've permanently deleted your Deleted recordings.

> Anyone know what would cause that?  These recordings were all from the 
> 2011/2012 timeframe, so it has been a while, but I am not sure why 
> that would be set (I am not sure what it even does to be honest).
>
> One nice thing - I got a free 1/2 TB of disk space back this morning :)

There's a race condition that occurs when you delete a huge number of 
shows and shut down mythbackend improperly (i.e. if you kill -KILL (or, 
even worse, if you kill it improperly with improper syntax, using kill 
-9)).  This usually happens because when you shut down mythbackend, even 
though it shuts down, it's still listed as a process until the file 
system finishes deleting all the shows that it had deleted.  Sometimes, 
users (or distro scripts?) get impatient and think mythbackend is hung 
or something and then use the nuclear option, which can leave some 
number of those just-deleted files in an intermediate/not-deleted 
state.  It's also possible that a power loss that occurs just after a 
huge delete could cause the problem (but, again, it's timing related and 
only will occur if the improper shutdown/power loss occurs right after a 
huge delete).

This will only occur on some systems when you delete a very large amount 
of recording data (huge total file size compared to the speed at which 
your file system deletes large files--where it would take many 10's of 
seconds or minutes to finish deleting) and when you shut down 
mythbackend at the wrong time and in the wrong way.

That said, the race condition will never occur if you enable slow 
deletes, so even if you don't need them, using slow deletes will prevent 
this problem from ever occurring.

Also, messing with the deletepending flag in the database can cause 
serious problems with mythbackend if you don't do it properly--so I 
highly recommend people don't mess with it.  You shouldn't have many (if 
any) recordings in limbo unless you frequently shut down mythbackend 
incorrectly after deleting lots of recordings.

The find_orphans.py script needs to be modified to detect these and 
clean them up, and I think Raymond has done so, but I don't think the 
changes ever made it into the wiki version of the script.  Eventually, 
the master backend will take care of these automatically on restart, but 
I haven't finished making the changes.

Mike