Make LXC container always live

[Marx]

Hi
I have 3 nodes Proxmox cluster. I've created one LXC container with the only responsibility being HAProxy for my cluster. It uses shared Ceph storage (each Proxmox node is Ceph node too). It has statically assigned IP, so it's irrelevant, on which node it runs - I can always access it via this IP. The problem is: when I shutdown node with this container, this container just die and nothing happens.
There are two situations:
- I gracefully shutdown Proxmox node - I expect live migration of container with zero downtime to another node (or at least restart on another node)
- Node is going down exceptionally (power outage, whatever) - I expect container to be restarted on another node
Do I miss any settings to make it working like this?

 

[H4R0]

Did you setup HA ?

Datacenter -> HA -> Add Container

 

[Marx]

yeah, good shot, just after posting I've made another search and together with "live migration" also "HA" came up. I've just set HA, Shutdown Policy, restarted a node, and it migrates great!
The only problem now is Ceph - sometimes OSD stops out of sudden and I need to manually start it. I can't understand why it happen, sometimes I even have to recreate Ceph on a node, because OSD doesn't start anymore. I use default time synchronisation, with default ntp servers from debian. I've read Ceph needs perfect time synchronization. My router can be NTP server, maybe I should use it instead of debian NTP servers?