[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Best Logfile size for XFS



Roger wrote:

>ok. saw something in the archives about logfile size asked within the
>past 2 days but it really didn't give any clues to this question.
>
>As the FAQ states, specifing an alternate logfile size (and also other
>options) at the time of mkfs, can increase performance.
>
>here's a quick layout of my partitions on (using an add-on Promise
>Ultra100 ide controller card):
>
>Filesystem            Size  Used Avail Use% Mounted on
>/dev/hde1             576M   69M  507M  12% /
>/dev/hde5             4.3G  2.3G  2.0G  53% /home
>/dev/hde6             207M  280k  206M   1% /tmp
>/dev/hde7             3.0G  2.5G  571M  82% /usr
>/dev/hde8             787M   84M  704M  11% /var
>/dev/hdf5             6.2G  2.8G  3.4G  45% /usr/src/RPM
>/dev/hdg6              16G   14G  3.3G  80% /mnt/win_c2
>/dev/hdg5              21G   33M   19G   1% /extra1
>/dev/hdf6             3.4G  3.3G  165M  96% /mnt/win_d2
>
>I'm mainly concerned about /dev/hdg5 (/extra1).  I'm using it for video
>capturing and am curious as to optimizing it for write performance.
>
>Logfile size recommendations? other options?
>
>Since they SGI touches on this in the "howto make an xfs", i'm sure
>there will be plenty of others asking the same question i am...but with
>raid & larger hdd's. ;-)
>
There is no hard and fast rule for this, a larger log is only really 
useful if you
are doing metadata intensive operations over extended periods of time and
we have found that more iclogs are just as useful (the logbugs=8 mount 
option).
I cannot remember right now, but mkfs may automatically make the
log bigger on large devices, of course large may be past the 2 Tbyte limit
on linux.

For write performance a larger log will not help, more iclogs might, but
not by much.

The other thing to consider with larger logs is that recovery after a crash
can take longer.

Steve