[mythtv-users] Odd: Storage Group is showing two different drives as the same MythTV drive

[Michael T. Dean]

On 07/25/2007 10:21 PM, Chris Pinkham wrote:
> I thought that this 5-minute rate
>   

It's a 5-second rate, right?

> was good enough, but maybe that needs tob e tightened down even more.

I think a 5-second rate is reasonable.  And, since the comparison is
only used if the total disk space is within 16KiB, most users probably
won't see the issue (assuming users tend to have differently-sized
partitions because of differently-sized drives or whatever).


However, if it is a 5-minute rate, I'd say it is too high (I'm sure
checking the list of directories takes less than 5 minutes ;).  For a
5-minute rate, Robert's 771MB/min would mean 2 like-sized partitions
with free space within 3855MB of each other would be treated as the same).

For the 5-second rate, on the other hand, the values seem reasonable. 
Robert's max recording rate of 771MB/min puts the magic difference value
at about 65MB/min (yeah, earlier I said 155--it wasn't my day for
math).  My system (with 2 HD-3000's in each of 2 backends with only
local storage on each) reports 277MB/min, which means filesystems have
to be within 23MB of each other.  I doubt we'd see many systems that get
much higher than 1GB/min--where the difference would have to be within 85MB.

But, since the time required to check all the directories is probably
quite a bit less than 5 seconds, we might be able to lower it.  However,
I'd think that a false positive test for uniqueness would be worse than
a false negative.  If we think 2 directories on the same filesystem are
actually on different filesystems, we'd be double-counting its free
space.  If we think 2 directories on different filesystems are actually
on the same, we'd under-estimate free space.

I guess which is worse is determined by what all else this affects
besides the status summary.  If it affects autoexpire, would it be worse
to fail to expire programs (and potentially run out of disk space) or to
expire a program when doing so may not have been necessary?  IMHO,
deleting a show that didn't have to be deleted just yet isn't too bad
because it will likely need to be deleted for the next recording,
anyway.  (I do realize that either of these situations would be
/extremely/ rare due to the confluence of circumstances required to
trigger them--i.e. users don't worry because this may happen once for
one or two users ever in the entire future of Myth.)

Even if it does affect recording-location decision making, I don't see
that as causing any problems.

If it only affects the summary, I guess it doesn't matter which way we
fail.  For that case, it's just a matter of preference.

The more Robert describes what was happening (recording and/or deletion
occuring when checking the backend status page), the more I'm amazed
that he even saw the issue.  Probably got a fortune cookie at lunch
yesterday that said, "Your timing will be impeccable today."

Mike