[mythtv] FW: Ticket #10541: SIGSEGV when EIT updating DB and channel change

Angela angela.schmid at wolke7.net
Sun Apr 15 18:21:50 UTC 2012


>On Sun, 2012-04-15 at 18:52 +0200, Angela wrote:
>> No crashes anymore. However I see mythbackend hanging, which I have seen
in
>> the past also without this patch. 
>
>The hangs you reported were the reason I didn't try to apply this before
>the 0.25 release.
>
>> During "EITCache: Wrote 14 modified entries of 252 for channel 1002 to
>> database" messages, I think mythbackend comes unresponsive, this
behaviour
>> has been seen before, also without this patch.
>> This time the recording has been canceled. In my previous comment, the
>> recording started 1.5 minutes later.
>> p.s. Why does MythTV not reuse prepared statements ? For each
>> insert/update/delete as far as I have seen, the SQL command is prepared
on
>> each call. Why not preparing it once and reuse it. Sure needs some
>> re-preparing when the connection has to be re-astablished. Writing 2000
EIT
>> records needs 3 Minutes, which is in my eyes very long. This is on a
Sandy
>> Bridge 2600K.
>
> I think the hangs and your slow database may be related. The non-reuse
> of prepared statements is historic. The code was written when Qt didn't
> support prepared statements. But even so your database seems to be
> running like it's writing to a floppy and not a hard disk. What is the
> formatting of the partition containing the database? If it is xfs or
> ext4 does adding nobarriers to the mount fix the hanging?

Dear Daniel,

Had this:
/dev/sda2 on / type ext4 (rw,errors=remount-ro)
UUID=8d6b1917-e239-43f0-a957-a5aa181e0255 /               ext4
errors=remount-ro 0       1

Changed to:
UUID=8d6b1917-e239-43f0-a957-a5aa181e0255 /               ext4
errors=remount-ro,nobarrier 0       1

Having now:
/dev/sda2 on / type ext4 (rw,errors=remount-ro,nobarrier)

By the way, I installed a new disk this week and reinstalled with 12.04. The
behavior seems to be the same. The new disks seems to be fine, moved music
data from an almost 10 year old PC over Gigabit LAN with 70MB/sec today.
However I had some warnings during a recording:

W TFWWrite ThreadedFileWriter.cpp:499 (DiskLoop)
TFW(/home/store/recordings/1009_20120415133000.mpg:34): write(57904) cnt 11
total 649352 -- took a long time, 1092 ms


This is MAGIC, wrote 2180 EIT messages within one second, 350 EIT messages
also in one second and 4000 EIT messages within 2 seconds.

I always use InnoDB, having nobarrier should be save for MySQL I think, in
case electricity fails. Losing parts of a recording does not matter. Any
other thoughts ?

Keeping an eye on the issue further.
Thanks, Angela








More information about the mythtv-dev mailing list