If an LXD-cluster with three nodes is overkill for you, a cold standby failover can easily be configured and works fine with two nodes. Be sure to configure WOL (Wake on Lan) for the cold standby node and ensure regular (skripted) container syncronisation or do it manually. Note also, container snapshots on the primary node belong to the container and are syncronised to the secondary node. You may want to prevent snapshots from being syncronised for speed and space with the option --instance-only.
One caveat of this model is that the controller node holding the primary database is not present. Thus ensure that the containers on the secondary node are stopped as the syncronised containers have identical IP's as those on the primary node which causes network issues. Stopping all containers on the secondary node at startup sounds strange, since this is one of th great advantages of LXD... but for this use case add @reboot sleep 60; lxc stop --all to your crontab so that all containers are stopped after cold-starting the secondary LXD-backup server.
Assuming you have two identical servers LXD1 (primary LXD-server) and LXD2 (secondary LXD-backup). Both servers should be known to eachother by adding them to remotes respectively. Needless to say that passwordless SSH between both servers should be configured.
Check remote servers:
lxc remote list
add remote server:
on LXD1
lxc remote add LXD2 <ip.address.of.server>:8443
add remote server:
on LXD2
lxc remote add LXD1 <ip.address.of.server>:8443
Copy container to LXD2
lxc copy CONTAINER LXD2: --stateless --refresh
Check containers on LXD2 server:
lxc list LXD2:
Copying or syncronising containers between servers can be scripted. Be sure to stop the containers to be syncronised on LXD2 server.
Example script:
#!/bin/bash
####################################################################
## Syncronise/copy containers and snapshots from LXD1 to LXD2 server
####################################################################
## Stop all containers on backup-server
ssh <ip.address.of.LXD2> 'lxc stop --all'
#copy container1
lxc copy CONTAINER1 LXD2: --instance-only --stateless --refresh ;
#copy container2
lxc copy CONTAINER2 LXD2: --stateless --refresh ;
exit
See lxc copy --help
lxc copy
Description:
Copy instances within or in between LXD servers
Transfer modes (--mode):
- pull: Target server pulls the data from the source server (source must listen on network)
- push: Source server pushes the data to the target server (target must listen on network)
- relay: The CLI connects to both source and server and proxies the data (both source and target must listen on network)
The pull transfer mode is the default as it is compatible with all LXD versions.
Usage:
lxc copy [<remote>:]<source>[/<snapshot>] [[<remote>:]<destination>] [flags]
Aliases:
copy, cp
Flags:
--allow-inconsistent Ignore copy errors for volatile files
-c, --config Config key/value to apply to the new instance
-d, --device New key/value to apply to a specific device
-e, --ephemeral Ephemeral instance
--instance-only Copy the instance without its snapshots
--mode Transfer mode. One of pull, push or relay (default "pull")
--no-profiles Create the instance with no profiles applied
-p, --profile Profile to apply to the new instance
--refresh Perform an incremental copy
--stateless Copy a stateful instance stateless
-s, --storage Storage pool name
--target Cluster member name
--target-project Copy to a project different from the source
Global Flags:
--debug Show all debug messages
--force-local Force using the local unix socket
-h, --help Print help
--project Override the source project
-q, --quiet Don't show progress information
--sub-commands Use with help or --help to view sub-commands
-v, --verbose Show all information messages
--version Print version number