[mythtv-users] Increasing ivtv and/or Myth buffering?

f-myth-users at media.mit.edu f-myth-users at media.mit.edu
Sun Aug 27 01:14:33 UTC 2006

[Crossposted to ivtv-users & mythtv-users because each might have
something useful to say about this.  ivtv 0.4.1, myth 0.18.1, several
PVR-250's, 2800+ AMD CPU, Ubuntu Breezy with ext3fs for OS & JFS for

I've noticed that operations such as reschedules in Myth sometimes
thrash the disk hard enough to cause IOBOUND errors and corrupted
recordings.  If the database has been recently optimized, the chances
of this are lower, but still significant (e.g., I most recently tripped
over this while changing a match condition in a rule that caused 30+
scheduled shows to all become unscheduled---I heard the disk thrash
for 5-10 seconds and both streams being recorded at the time got
IOBOUND errors and had many seconds missing during that interval.)
I also see a suspicious correlation between IOBOUNDs in my logs and
other scheduler runs, including after mythfilldatabase.

I'd -really- like to avoid having to put the Myth database (and
perhaps the entire OS) on a second drive in the machine -just- to
avoid what shouldn't be happening in the first place---not only would
I need to buy another drive, but I'd rather not generate the additional
heat and burn the additional power.  This seems entirely related to head
motion and not at all related to total throughput; I can record 5
simultaneous streams on this machine, or copy data at full rate from
its 100baseT network, without causing glitches like this.

Is there some way of increasing the amount of buffering uses by ivtv
so it's more tolerant of hangups waiting to get to the disk?  Or some
way of increasing the amount of buffering that Myth uses after it's
read data from ivtv but before it's written it to disk?  I'd like to
double or triple the total amount of buffering between the tuner and
the disk, if possible.  (I'm assuming that the answer is the same even
with these oldish ivtv & myth versions, but if I can only increase
buffering in some newer release, please let me know.)

[ISTR some sort of number-of-buffers argument to ivtvctl, but can't
find it now that I'm looking, and web searches haven't been helpful.]

The machine in question has 512meg of RAM in it, but could trivially
have that doubled, since I have another 512meg stick freed up from
another machine that's simply been waiting for the next powerdown.
According to top, RAM usage is currently ~510meg used, 6meg free, 500k
buffers, 172meg cached (and 200meg used in swap), so I'm guessing that
increasing the RAM if I increase buffering would be a reasonable idea
in any event.

Thanks much...

More information about the mythtv-users mailing list