Usually i use kernel 2.6.9 , but now i have to move to 2.6.18 (2.6.18-8.1.15.el5.028stab047.1PAE) and i found out that "vzctl stop" - hangs for exactly 2 mins (start is ok and runs ok )
I found out that it happens only with VEs which use loop mounts ( mount -o loop,ro) - i use them for sharing data among VEs.
I made strace during stop (it helped me once when stop hanged cause couldnt find "halt" binary ), but wasnt able to find out the reason . Anyway i attached strace.txt .
Also i noticed that it doesnt umount clean loop devices .
On prvious kernel i had in .umount
umount /vz/root/$VEID/vps_static
umount /vz/root/$VEID/vps_libs
On new kernel i get error
[dagr@ssw2-ca shared_vps]$ date;sudo vzctl stop 1051206;date
Wed Nov 14 14:05:40 GMT 2007
Stopping VE ...
VE was stopped
umount
umount: /u04/vz/shared_vps/php5_base/vps_libs_loop: not mounted
umount: /u04/vz/shared_vps/php5_base/vps_libs_loop: not mounted
Error executing umount script /etc/vz/conf/1051206.umount
Wed Nov 14 14:07:41 GMT 2007
And i get 1 loop device not detached (from /vz/root/$VEID/vps_libs mount )
[dagr@ssw2-ca private]$ ps -ef| grep loop
root 601 1 0 14:04 ? 00:00:00 [loop1]
[dagr@ssw2-ca private]$ sudo losetup -d /dev/loop1
I thought that may be openvz on new kernel somehow handles loop umounts by itself , but if comment .umount file , it leaves all 2 mounts.
(vzctl-3.0.18-1 , ovzkernel-PAE-2.6.18-8.1.15.el5.028stab047.1)
So 2 questions.
1. Whats wrong with stop ?
2. Whats wrong with loop device umount ?