[mythtv-users] mythbackend crashes every 6-7 hours

James Crow crow.jamesm at gmail.com
Mon Nov 1 21:00:54 UTC 2010


On Mon, Nov 1, 2010 at 4:12 PM, Gavin Hurlbut <gjhurlbu at gmail.com> wrote:
> On Mon, Nov 1, 2010 at 12:32 PM, Rob Verduijn <rob.verduijn at gmail.com> wrote:
>> system-data2 /data2 443Gb XFS
>> system-data2 is used for the recordings.
>> There is 59Gb free space left on that partition
>> No redundancy has been provided because I  don't care about the local data.
>
> OK, by my calculation, this is about 87% full.  XFS is known to have
> issues with filesystems that are approaching full (bad fragmentation,
> IIRC).
>
> xfs_db -c frag -r /dev/whatever/system-data2
> xfs_db -c freesp -r /dev/whatever/system-data2
>
> The first shows fragmentation in the used drive space, and the second
> shows the fragmentation in the free drive space (which is more likely
> to be the issue at hand).
>

I do no think running XFS drives at nearly full is any problem. I have
three spindles dedicated to recordings and all three run XFS at >95%
full. I do run defrag for a couple of hours each night on them and I
have them set to allocsize=512M. That helps some with the
fragmentation, but even when they are 80% or more fragmented I do not
see crashes like Rob is seeing. I have my system set to auto-expire
vice delete recordings so the drives stay full.

I would recommend anyone run the xfs_db commands above and set their
drives to defrag when not recording. Amount of time required to defrag
depends on number, size, and frequency of recordings among other
things.

Cheers,
James


More information about the mythtv-users mailing list