ub16.04 lxc container hang on start (5.0beta)

[gaz]

Hi all,
I hope this isn't a newbie mistake.
I've got an installation of 5.0beta, updated current to current. This was a direct upgrade from 4.4
When I create lxc containers using the stock ubuntu 16.04 image and run them, I get no response for about the first 5 minutes of running. No CPU is shown as used, and the container becomes pingable (IP by DHCP). SSH connects are refused. Connecting to the console gives no response.
At about the 5-minute uptime mark, the CPU use goes up for a few seconds and the console starts responding.

This only happens with ubuntu and not debian or centos images.

Other details: root is a zfs zvol, 10GB
cpu - 1 core, no limits
network - bridged
unprivileged container

Is there something special about systemd in ubuntu that's causing problems? This seems like a timeout, but I can't find anything in the syslogs of the container.

 

[gaz]

So, it seems to be the IPV4 DHCP setting. If I run with a static IP, the LXC image boots normally. If I ask for IPV4 DHCP, it hangs 5 minutes on boot, though it gets the lease.
After a little more careful testing, I think this may be happening with more than just ubuntu - seems to also affect the debian pre-built container

 

[dcsapak]

do you maybe have set ipv6 also to dhcp without having an ipv6 dhcp server?

 

[gaz]

I do have dhcp6 on the network. It's a home network with DHCP 4 & 6 served off of the home router ASUS RT-AC68.
I tried running the containers with either static and DHCP IPV6 - no difference. The only variable that had an effect was ipv4 address (static vs DHCP). Perhaps there's some peculiarity with my DHCP server.
I found a similar problem previously reported, but it could not be replicated.
http://pve.proxmox.com/pipermail/pve-user/2016-May/155798.html

I will try to reproduce this in a more controlled environment with pfsense running in a VM serving up the DHCP to the container.
It's not a deal-breaker for me, since I can/should assign the IP's myself anyway.

 

[gaz]

I've confirmed that the DHCP / LXC container DHCP hang issue does NOT happen when using the DHCP server from pfsense 2.3.3 (both on the same vmbridge).
This must be a strange corner case.