Skip to contentSkip to conversation

gentoo linux, kernel 3.10.31 mount options bug?

DISCUSSION Marko Weber|8000 3 participants 4 messages Feb 23, 2014 – Feb 25, 2014

lore
Share & tools
lore
#0 lore raw

hello list,

i have a question about mount options.

at first my fstab:

/dev/md0                /boot           xfs             noauto,noatime  1 2
/dev/md2                /               xfs             noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep      0 1
/dev/md1                /tmp            xfs             noexec,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep 
      0 1
/dev/VolGroup01/var-www /var/www        xfs             noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep      0 1
/dev/VolGroup01/var-log /var/log        xfs             noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep      0 1
/dev/VolGroup01/var-lib-mysql   /var/lib/mysql  xfs             noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep      0 1
/dev/sda5               none            swap            sw,pri=1         
       0 0
/dev/sdb5               none            swap            sw,pri=2         
       0 0

(YES!, normally i use Label or UUID..... but this is only a test machine.....)

AFTER reboot "mount" shows this:

rootfs on / type rootfs (rw)
/dev/root on / type xfs (rw,noatime,nodiratime,attr2,nobarrier,inode64,noquota)
/dev/md1 on /tmp type xfs (rw,noexec,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--www on /var/www type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--log on /var/log type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--lib--mysql on /var/lib/mysql type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep)

on /dev/root "discrad" and some other are not shown

when i do a "mount -o remount /"

"mount" shows this:

rootfs on / type rootfs (rw)
/dev/root on / type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep)
/dev/md1 on /tmp type xfs (rw,noexec,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--www on /var/www type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--log on /var/log type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/mapper/VolGroup01-var--lib--mysql on /var/lib/mysql type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep)

now,all options from /etc/fstab are listed.
Is this XFS related?

Kernel 3.10.31
Grub2

in the /etc/grub.d/40_custom i have set:

menuentry "gentoo linux 3.10.31" {
insmod xfs
insmod mdraid09
insmod mdraid1x
insmod lvm
set root=(hd0,2)
linux /bzImage-3.10.31 root=/dev/md2 elevator=cfq clocksource=hpet cgroup_disable=memory }

If you need any more info, let me know.

thnaks for any tipps or answers

marko

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#1 lore raw
Quoted 10 lines · On 2/23/14, 3:37 AM, Marko Weber|8000 wrote:

hello list,

i have a question about mount options.

at first my fstab:

/dev/md0                /boot           xfs             noauto,noatime  1 2
/dev/md2                /               xfs             noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep      0 1

I guess I won't ask why you decided to tweak all those knobs...

<snip>

Quoted 7 lines

(YES!, normally i use Label or UUID..... but this is only a test machine.....)

AFTER reboot "mount" shows this:

rootfs on / type rootfs (rw)
/dev/root on / type xfs (rw,noatime,nodiratime,attr2,nobarrier,inode64,noquota)

<snip>

default mount options

Quoted 4 lines

on /dev/root "discrad" and some other are not shown

Quoted 8 lines

when i do a "mount -o remount /"

"mount" shows this:

rootfs on / type rootfs (rw)
/dev/root on / type xfs (rw,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep) /dev/md1 on /tmp type xfs (rw,noexec,noatime,nodiratime,nobarrier,largeio,discard,logbsize=256k,noikeep)

tweaked options.

<snip>

Quoted 3 lines

now,all options from /etc/fstab are listed.
Is this XFS related?

Not really. Only some xfs options are remountable, so unless you originally mounted with, say, "largeio" the option won't be in effect. Mount -o remount won't enable it. So if you really want all those options enabled on your root fs, you'll need to have them in your initrd/initramfs so that the original mount of the root fs at boot time includes them.

Also, mount may be looking at /etc/mtab rather than /proc/mounts. /proc/mounts contains the true state of the fileystem w.r.t. mount options. /etc/mtab is managed by mount, and knows nothing about what the filesystem actually did:

[root@sandeen ~]# mount -o remount,notanoption /home

[root@sandeen ~]# mount | grep home
/dev/md2 on /home type xfs (rw,notanoption)

[root@sandeen ~]# grep home /proc/mounts
/dev/md2 /home xfs rw,seclabel,relatime,attr2,delaylog,noquota 0 0

-Eric

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#2 lore raw
In reply to Marko Weber|8000

On 2/23/2014 3:37 AM, Marko Weber|8000 wrote:
...

Quoted 1 line

linux /bzImage-3.10.31 root=/dev/md2 elevator=cfq clocksource=hpet

                                       ^^^^^^^^^^^^

cfq tends to defeat much of the parallelism in XFS, decreasing throughput substantially. This is documented in the XFS FAQ and has been discussed here many times in the past. It has been recommended for a few years now that XFS not be used with the cfq elevator. Use deadline with md arrays on plain HBAs and noop on SSDs or any device with [F|B]BWC, i.e. RAID HBA or SAN controller.

If you're using cfq to allow shaping of per process IO with control groups, simply using cfq alone may slow down XFS throughput to the point that you don't need to bother with control group optimizations.

Signature · 6 lines
Stan

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
#3 lore raw

Thanks for the short and clear answer stan.

Marko

Quoted 20 lines · Stan Hoeppner

On 2/23/2014 3:37 AM, Marko Weber|8000 wrote:
...

Quoted 1 line

linux /bzImage-3.10.31 root=/dev/md2 elevator=cfq clocksource=hpet

                                      ^^^^^^^^^^^^

cfq tends to defeat much of the parallelism in XFS, decreasing throughput substantially. This is documented in the XFS FAQ and has been discussed here many times in the past. It has been recommended for
a few years now that XFS not be used with the cfq elevator. Use deadline with md arrays on plain HBAs and noop on SSDs or any device with [F|B]BWC, i.e. RAID HBA or SAN controller.

If you're using cfq to allow shaping of per process IO with control groups, simply using cfq alone may slow down XFS throughput to the point
that you don't need to bother with control group optimizations.

Signature · 1 line
Stan
Signature · 1 line
Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs