[mythtv-users] mythfilldatabase vs DataDirect lossage

f-myth-users at media.mit.edu f-myth-users at media.mit.edu
Mon Dec 18 22:56:53 UTC 2006


A couple of days ago, DataDirect was -very- slow, causing my
mythfilldatabase run to take more than 4 hours to complete (and it
timed out on most of the fetches).  This in turn caused one of my
lineups to say NO DATA in MythWeb for all of that day, and indeed
all scheduled recordings for that lineup for that day unscheduled
themselves, which caused me to miss several recordings.  (After all,
they -had- been correct, and I was away from Myth all day, and then
when I came back, the current day's listings were all NO DATA and
recordings that had been scheduled for that day for almost two weeks
hadn't happened.)

DD was quite slow today, too, although apparently not enough to
-totally- time out.  (Although several attempts got two successive
"401 unauthorized" errors instead of the typical 401/200 pair, so
presumably those fetches failed, too, and the average transfer rates
were 5-10x below typical.  No, it's not at my end.)

It seems like a really bad idea for mfdb to -ever- replace good data
(obtained 13 days before, and 0-1 days before, via the normal behavior
of mfdb) with "NO DATA" the day of the recording because of a timeout.
Do modern (.20 or SVN) mfdb's do this, or was this peculiar to the
18/19 era?  (I've seen some patches go by on -dev about mfdb behavior
in certain cases like this, but I thought they were mostly for what
happens near the ragged edge 13 days out, not the current day).

If current versions of mfdb -do- simply delete existing scheduling data
because they didn't happen to get up-to-date data (for the current day
and "tomorrow"), I'd like to advocate that this get changed---it's
-far- better to just carry on for (especially!) the current day just
in case the fetch got screwed up than to -guarantee- that a fetch
failure will destroy the recording schedule---especially "today's"
schedule, which is by definition the least likely to be caught in time.

For mfdb to delete good data because wget timed out means that no
matter how recently one has looked at the upcoming-recordings
schedule, it's still possible to get screwed, since mfdb may run
at any time (if you're following DD's recommendations) which means
you never know when your good schedule might get trashed without
warning.  (If I have to babysit Myth every few minutes/hours to make
sure it's not about to delete all my scheduled recordings, that pretty
much destroys the whole utility of a PVR.)


More information about the mythtv-users mailing list