[OmniOS-discuss] LX for OmniOS update

[Jim Klimov]

10 августа 2016 г. 1:20:19 CEST, Michael Rasmussen <mir at miras.org> пишет:
>On Wed, 10 Aug 2016 00:11:39 +0200
>Michael Rasmussen <mir at miras.org> wrote:
>
>> I have taken the liberty to extend your script to handle gzipped
>> tarballs of a root filesystem using the argument -t. With the stuff I
>> have added I was able to use the script to install a lxc prepared
>> tarball of Debian stable with your install script;-)
>> 
>By the way. Linux LXC prepared tarballs can be downloaded here:
>http://download.proxmox.com/images/system/
>
>The one I used is this one:
>http://download.proxmox.com/images/system/debian-8.0-standard_8.0-1_amd64.tar.gz

Michael, thanks for posting the script. There was a typo to fix (un-tarring goes into ZONEPATH not ZONEPATH_DS, and a 'gtar' worked better for my tarball).

One Linux-based project I work with builds OS images into Debian 8 tarballs that can run as the rootfs of an LXC container or via chroot, either way without a kernel of its own. It seemed natural to try it out as a zone :-)

The systemd startup reported failures of mounting /tmp (not permitted). Kernel device manager (udevd) also failed (maybe because it can't create /dev/xconsole fifo - 'not implemented'). And Load/Save random seed (/dev/urandom - permission denied). Also says something about 'Failed to attach new connection bus to event loop: Operation not permitted' many times.

Network automatic setup failed, but subsequent manual ifconfig was ok (I can curl to a server in this LAN segment).

Logging does not work (no dmesg, no journalctl for failed services, although there are some lines in common `journalctl -l`).

A mysql (mariadb) server has started and initialized its database ;) Its consumers, even scripted ones (a schema uploader/verifier), did not. Many units report that they failed to run their 'start' tasks with 'Operation not permitted' (maybe passed up from bus-connection attempts).

Requesting `systemctl status` or `systemctl restart` of services fails (Connection reset by peer). Not sure what it connects to - in our project OS image there is no D-Bus, so in true Linux a simple user can not use systemd tools, but for a root (or via sudo) there were no issues using it.

Login works both on system console (from zlogin -C) and from GZ (simply zlogin). `poweroff` from inside guest zone sends a message that it will go down, and does nothing. `zoneadm halt` from GZ works.

Setting kernel version in zone attrs to '3.18.0' did not seem to break nor fix anything, but the value is seen in `uname -a`.

Hope these ramblings help,
Jim Klimov


--
Typos courtesy of K-9 Mail on my Samsung Android