| 1 |
On Tue, Dec 01, 2015 at 12:24:05AM +0000, Robin H. Johnson wrote: |
| 2 |
> A quick grep in the tree shows the following existing cases (wrapped for line |
| 3 |
> length). |
| 4 |
> |
| 5 |
> $ git grep -E \ |
| 6 |
> 'keyword .*(lxc|systemd-nspawn|vserver|prefix|openvz|uml|jail|xen)' |
| 7 |
> sys-apps/kmod/files/kmod-static-nodes-r1: |
| 8 |
> keyword -lxc -systemd-nspawn |
| 9 |
> sys-fs/eudev/files/udev-postmount: |
| 10 |
> keyword -vserver -lxc |
| 11 |
> sys-fs/zfs/files/zfs-0.6.1-gentoo-openrc-dependencies.patch: |
| 12 |
> keyword -lxc -openvz -prefix -vserver |
| 13 |
> sys-fs/zfs/files/zfs-0.6.5-fix-openrc-scripts.patch: |
| 14 |
> keyword -lxc -openvz -prefix -vserver |
| 15 |
> sys-power/nut/files/nut.powerfail.initd: |
| 16 |
> keyword -jail -lxc -openvz -prefix -uml -vserver -xenu -timeout |
| 17 |
> |
| 18 |
> The only one of those I can speak knowledgeably about is sys-power/nut. |
| 19 |
> That script signals the UPS to cut the power, and is intended to be the very |
| 20 |
> very last shutdown script ever called on a box (so it has a sleep forever bit |
| 21 |
> at the end). |
| 22 |
> |
| 23 |
> Does it make sense in a container? That depends on how the container is |
| 24 |
> configured... it's got the right details for the container parent's UPS, then |
| 25 |
> you certainly don't want it to fire. If it had details to tell the parent to |
| 26 |
> just terminate it, then it would be correct. |
| 27 |
> |
| 28 |
> Rather than replacing all of the system-specific keywords, can we make |
| 29 |
> 'container' into an alias that expands to the full list of known container |
| 30 |
> types? That way, if there is an init script that is specific to some container |
| 31 |
> type, it would still be usable. |
| 32 |
|
| 33 |
This is the approach I would personally be more comfortable with. |
| 34 |
|
| 35 |
William |
Archives