Package: aoetools
Version: 30-3
Severity: normal
Tags: patch
There are two issues with /etc/init.d/aoetools. One is on startup, the
second on stop.
I have a patch - hopefully attached to this report. I didn't bother to send
a patch for /etc/default/aoetools, where i've only defined AOE_TMOUT to be
30.
The startup problem
~~~~~~~~~~~~~~~~~~~
Scenario:
- ocfs2 filesystem
- on an aoe device
- accessed via the bond0 (round-robin) interface
the relevant line in fstab is
/dev/etherd/e104.0 /srv ocfs2 defaults,_netdev 0 0
What happens with the original setup is that
1 bond0 is brought up with ifup
2 the aoe module is loaded and /dev/etherd/discover shows up
3 /etc/init.d/aoetools tries to mount /srv, but /dev/etherd/e104.0 doesn't
yet exist
4 the ocfs2 cluster gets initialized and /etc/init.d/ocfs2 tries to mount
/srv, but the block device is still not there
5 link comes up on the slaves of bond0
6 /dev/etherd/e104.0 is discovered (a few long seconds too late)
My solution:
- keep list of block devices rather than mountpoints in $waitaoe
- in a time-limited loop, test for the presence of a block device before
trying to mount it or sleep one second otherwise
- the above loop is limited to $AOE_TMOUT (defined in /etc/default/aoetools)
seconds
The stop problem
~~~~~~~~~~~~~~~~
In my setup, "/etc/init.d/ocfs2 stop" will umount the AOE-backed filesystem,
so when "/etc/init.d/aoetools stop" tries to umount it (again!), the umount
command fails, and because of the "set -e", the script exits immediately,
skipping the LVM/vgchange part and the removing of the module.
My approach was to check if the block device is mounted before trying to
umount.
I hope my changes to the init script don't break anything else. In
particular, I haven't tested any LVM on AOE on this system.
Regards,
adc
-- System Information:
Debian Release: 5.0.3
APT prefers proposed-updates
APT policy: (500, 'proposed-updates'), (500, 'stable')
Architecture: amd64 (x86_64)
Kernel: Linux 2.6.26-2-openvz-amd64 (SMP w/4 CPU cores)
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
Shell: /bin/sh linked to /bin/bash
Versions of packages aoetools depends on:
ii libc6 2.7-18lenny1 GNU C Library: Shared libraries
ii lsb-base 3.2-20 Linux Standard Base 3.2 init scrip
aoetools recommends no packages.
aoetools suggests no packages.
-- no debconf information
--- aoetools.debian 2009-12-03 11:18:57.000000000 -0500
+++ aoetools 2009-12-03 12:09:03.000000000 -0500
@@ -61,7 +61,7 @@
continue
;;
/dev/etherd/*)
- waitaoe="$waitaoe $MTPT"
+ waitaoe="$waitaoe $DEV"
esac
done
@@ -130,10 +130,28 @@
if [ -n "$waitaoe" ]
then
- for mountpt in $waitaoe; do
- echo "Mounting $mountpt..."
- mount $mountpt
- #log_action_begin_msg "Waiting for $mountpt."
+ for dev in $waitaoe; do
+ i=0
+ echo -n "Waiting for and mounting $dev:"
+ while test $i -lt $AOE_TMOUT
+ do
+ if test -b $dev
+ then
+ echo -n " ready,"
+ mount $dev && echo " mounted." || echo ""
+ break
+ else
+ aoe-discover
+ echo -n "."
+ sleep 1
+ let i=$i+1
+ fi
+ done
+ if test $i -eq $AOE_TMOUT
+ then
+ echo " TIMEOUT!!"
+ fi
+ #log_action_begin_msg "Waiting for $dev."
done
else
echo "Nothing to mount."
@@ -145,9 +163,8 @@
if [ -n "$waitaoe" ]
then
- for mountpt in $waitaoe; do
- echo "Unmounting $mountpt..."
- umount $mountpt
+ for dev in $waitaoe; do
+ grep ^$dev /proc/mounts && echo "Unmounting $dev..." && umount $dev
done
fi
@@ -171,7 +188,7 @@
;;
stop)
- echo -n "Stopping $DESC: "
+ echo "Stopping $DESC: "
do_stop
