LinHES Forums
http://forum.linhes.org/

Filesystem performance boost, R5.5 install
http://forum.linhes.org/viewtopic.php?f=3&t=18473		 Page 3 of 3
Author:  graysky [ Mon Jul 21, 2008 1:25 pm ]
Post subject: 
nharris wrote:
How about using the "noatime" mount option in /etc/fstab (to stop file access times from being written)? Does this also help performance with XFS like it does EXT3?


If you look back on the 1st page of this thread:

tjc wrote:
BTW - atime is access time for the files, and the "noatime" option disables tracking it. It's a silly micro optimization unless you're running something with large numbers of smallish files and lots of parallel access (say like an image or news server).
Author:  graysky [ Thu Jul 24, 2008 1:18 pm ]
Post subject: 
About this whole atime, noatime discussion. It seems that R5.5 uses the relatime switch by default anyway:

Code:
$ cat /proc/mounts
rootfs / rootfs rw 0 0
/dev/hda1 / ext3 rw,relatime,data=ordered 0 0
none /proc proc rw,relatime 0 0
none /sys sysfs rw,relatime 0 0
/dev/hda1 /dev/.static/dev ext3 rw,relatime,data=ordered 0 0
tmpfs /dev tmpfs rw,relatime 0 0
usbfs /proc/bus/usb usbfs rw,relatime 0 0
tmpfs /dev/shm tmpfs rw,relatime 0 0
nfsd /proc/fs/nfsd nfsd rw,relatime 0 0
/dev/hda3 /myth xfs rw,relatime,allocsize=524288k 0 0
rpc_pipefs /var/lib/nfs/rpc_pipefs rpc_pipefs rw,relatime 0 0
Author:  nharris [ Thu Aug 14, 2008 7:33 am ]
Post subject: 
Just wanted to add my 2 cents.

One of XFS's features is online filesystem defragmentation. xfs_fsr is supposed to be safe for online and in-use file systems. I'm not saying that going into single user mode first is not a safe move. I just don't believe that it is necessary with a production tested filesystem like XFS. I put the xfs_fsr in my cron.daily and it runs every night to keep my fragmentation near 0%.
Author:  auttila [ Mon Nov 03, 2008 1:37 pm ]
Post subject: 
nharris wrote:
xfs_fsr is supposed to be safe for online and in-use file systems.


I just switched to XFS, andyep, after trasferring all the files back to /myth, the fragmentation got up to 99.8%. So little defragging is needed. I added the xfs_fsr to cron just like above, and hopefully it will take care of the problem.

However, I'm a little sceptic for it affecting the performance. My MythBox does a lot transconding all the time, and I have my doubts for defragmenting while transcoding. So the safest thing would be wait for an idle period to do the defragging.

Problem is, there's no way of telling when the mythbox is idle for the next 7200 seconds. Then a friend of mine came up with an idea to run xfs_fsr as a userjob for a few minutes for every recording.

That way the backend machine would not be doing anything else during defragmentation, right? Can someone wiser than me tell me am I going to nuke the whole system by doing that?
Author:  nharris [ Mon Nov 03, 2008 1:42 pm ]
Post subject: 
There is an idle.sh script which may help you. Search on suspend, mythwelcome and such for more details.

On a side note, I have been running with the cron script for a while now with no problems. However, my box is not super busy.
Page 3 of 3 All times are UTC - 6 hours
Powered by phpBB® Forum Software © phpBB Group
http://www.phpbb.com/