Bug#595926: invoke-rc.d lxc stop should provide a way to gracefully terminate containers
On Sat, 25 Sep 2010 11:13:22 +0200 Mike Hommey m...@glandium.org wrote: On Tue, Sep 07, 2010 at 03:13:26PM +0400, Konstantin Khomoutov wrote: Currently, LXC init script is able to start the configured set of containers (which is very convenient) but to stop them it uses lxc-stop which merely kills the set of processes belonging to a container. When a container has its own rootfs with a system created using debootstrap or similar means, it is expected that the stop action should be similar to start, that is, something like `shutdown -h +0` or `telinit 0` should be executed inside such containers, and then they should be waited upon to terminate. [...] Without lxc-attach, you can still use a workaround, provided your containers have a proper /dev/initctl and /sbin/telinit works: You can chroot $container_root telinit 0 Just verified on a test container -- it works. Thanks, Mike! While I think that having lxc-attach working is a must as Debian ships an LXC-enabled kernel, for now this workaround can be implemented in the lxc package as is. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#595926: invoke-rc.d lxc stop should provide a way to gracefully terminate containers
On Tue, Sep 07, 2010 at 03:13:26PM +0400, Konstantin Khomoutov wrote: Package: lxc Version: 0.7.2-1 Severity: normal Currently, LXC init script is able to start the configured set of containers (which is very convenient) but to stop them it uses lxc-stop which merely kills the set of processes belonging to a container. When a container has its own rootfs with a system created using debootstrap or similar means, it is expected that the stop action should be similar to start, that is, something like `shutdown -h +0` or `telinit 0` should be executed inside such containers, and then they should be waited upon to terminate. As I understand, the following sequence should work: for cont in $CONTAINERS; do lxc-attach -n $cont 'shutdown -h +0' \ lxc-wait -n $cont -s STOPPED \ lxc-stop -n $cont done The actual code can become more involved as ideally the init script should wait for some grace period for the containers to shut down and then use lxc-stop on the remaining containers forcibly. I assume the implementation details are open for discussion. Also, since containers are not required to virtualise a whole OS on its own rootfs and may be used to host just one process or a bunch of processes, there should be a way to mark container explicitly as requiring a graceful shutdown. Another possibility would be to specify a command to issue inside a container to shut it down; an empty command (the default) could signalize no graceful shutdown is required. This method would put more burden on the administrator but is more flexible. In any case, implementing this functionality depends on making lxc-attach working, and I filed bug #595920 for this reason. Without lxc-attach, you can still use a workaround, provided your containers have a proper /dev/initctl and /sbin/telinit works: You can chroot $container_root telinit 0 Mike -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#595926: invoke-rc.d lxc stop should provide a way to gracefully terminate containers
Package: lxc Version: 0.7.2-1 Severity: normal Currently, LXC init script is able to start the configured set of containers (which is very convenient) but to stop them it uses lxc-stop which merely kills the set of processes belonging to a container. When a container has its own rootfs with a system created using debootstrap or similar means, it is expected that the stop action should be similar to start, that is, something like `shutdown -h +0` or `telinit 0` should be executed inside such containers, and then they should be waited upon to terminate. As I understand, the following sequence should work: for cont in $CONTAINERS; do lxc-attach -n $cont 'shutdown -h +0' \ lxc-wait -n $cont -s STOPPED \ lxc-stop -n $cont done The actual code can become more involved as ideally the init script should wait for some grace period for the containers to shut down and then use lxc-stop on the remaining containers forcibly. I assume the implementation details are open for discussion. Also, since containers are not required to virtualise a whole OS on its own rootfs and may be used to host just one process or a bunch of processes, there should be a way to mark container explicitly as requiring a graceful shutdown. Another possibility would be to specify a command to issue inside a container to shut it down; an empty command (the default) could signalize no graceful shutdown is required. This method would put more burden on the administrator but is more flexible. In any case, implementing this functionality depends on making lxc-attach working, and I filed bug #595920 for this reason. -- System Information: Debian Release: squeeze/sid APT prefers testing APT policy: (500, 'testing') Architecture: i386 (i686) Kernel: Linux 2.6.32-5-686 (SMP w/4 CPU cores) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages lxc depends on: ii libc6 2.11.2-2 Embedded GNU C Library: Shared lib ii libcap2 1:2.19-3 support for getting/setting POSIX. Versions of packages lxc recommends: ii libcap2-bin 1:2.19-3 basic utility programs for using c lxc suggests no packages. -- Configuration Files: /etc/default/lxc changed: RUN=yes CONF_DIR=/etc/lxc CONTAINERS=bender -- no debconf information -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
