some containers are in pending state after each openstack deployment

[Aymen Frikha]

juju 2.0.2
maas 2.1.2
After each deployment, we always have 2 or more containers in pending state with juju status.
The container is up and running on the host and get its ip addresses. These are cloud init logs, cloud-init-output and syslog from the host:

https://pastebin.canonical.com/176082/

[Aymen Frikha (aym-frikha)]

log for container

[Aymen Frikha (aym-frikha)]

Dup of: https://bugs.launchpad.net/juju/+bug/1602192

[Billy Olsen (billy-olsen)]

This bug should no longer be duplicated to the other bug as that bug affects the local provider and this bug affects the MAAS provider. After this was discussed, it turns out that there is no intention of fixing this very specific case within Juju (LXD containers on MAAS provider) as those tunables should be left to LXD or the base OS.

A known work-around is to edit the MAAS preseed file and add necessary lines to adjust system tunables. For example, add the following lines to the end of /etc/maas/preseeds/curtin_userdata:

sysctl_fs_01: ["curtin", "in-target", "--", "sh", "-c", "/bin/echo 'fs.inotify.max_queued_events = 1048576' >> /etc/sysctl.d/15-fs.conf"]
sysctl_fs_02: ["curtin", "in-target", "--", "sh", "-c", "/bin/echo 'fs.inotify.max_user_instances = 1048576' >> /etc/sysctl.d/15-fs.conf"]
sysctl_fs_03: ["curtin", "in-target", "--", "sh", "-c", "/bin/echo 'fs.inotify.max_user_watches = 1048576' >> /etc/sysctl.d/15-fs.conf"]
sysctl_vm_01: ["curtin", "in-target", "--", "sh", "-c", "/bin/echo 'vm.max_map_count = 262144' >> /etc/sysctl.d/15-vm.conf"]