[mythtv-users] Issues with 0.25

[Tom Lichti]

On Tue, Mar 6, 2012 at 1:15 PM, Gavin Hurlbut <gjhurlbu at gmail.com> wrote:
> On Tue, Mar 6, 2012 at 7:02 AM, Scott and Nicole Harris
> <snharris99 at live.com> wrote:
>> itself is an issue]).  The part that seems to be being overlooked here, is
>> that users experiencing these bad recordings in 0.25 can remedy the
>> situation by rolling back to 0.24 (myself included), suggesting that the
>> issue is in 0.25, not hardware or corrupt data.  The issue only seems to
>> affect HDHR users and only when performing more than one recording at a time
>> (as near as I can tell by the "me too's" in this thread).
>
> If the database is not corrupt, then I'd suggest using mysqltuner.pl
> and see if there isn't something you need to tweak in my.cnf.
>
> The one huge difference between 0.24 and 0.25 is that in 0.25 we log
> to the database by default.  This adds a non-insignificant additional
> load to mysql, but if it is setup properly, there's no reason it
> should cause alarm.  To see if that's the problem, start the frontend
> and backend with --nodblog and see if it changed.
>
>> Other than max_connections being set to suitably (much higher) amount, what
>> other specific settings should we be looking at?
>
> Sizes of caches and the like, primarily.
>
> wget mysqltuner.pl
>
>> Perhaps the total revamping of the MySQL thread pool that now requires such
>> a large increase in max_connections is contributing to the high CPU usage
>> with multiple recordings?
>
> Doubtful.  I often have 6 recordings at once, additional to a MythTV
> compile and 4 flat out commflag runs, and I don't see any issues.  And
> my database is on ext4 with barriers enabled.  Granted, I'm using an
> i7-860, so I have 4 hyperthreaded cores (so 8 CPUs as seen by Linux).
>
>> Is using xfs for recording drives over ex4 a potential issue?
>
> Not likely.

Another of those 'me too' posts, I see significant CPU usage with HDHR
recordings now, and 3 or more concurrent recordings will bring my
backend to it's knees. I have nodblog enabled, the only thing
mysqltuner.pl complains about is the size of the table cache, which
shouldn't have any effect on this. My backend is a 2x dual core AMD
2.6GHz, so 4 cores, 8GB of RAM, and it will run two cores at close to
100% while recording from the HDHR's, which seems very high for
something that is literally a pass through to disk. My OS is on one
drive, DB on another (SSD) drive, and recordings are on a third drive.
I *may* be able to re-architect my backend, in terms of storage, but I
can't see why I should have to?

Tom