It would be great if we could get a clustering video that covers a webserver/db server combo. Including data replication across the cluster, ingress, failover. Best practices on container disk partitioning on the db etc.
Perhaps you could throw in backup/recovery of one of the containers.
We’ve got some videos on clustering, although they, understandably, focus on LXD itself rather than specific application workloads.
Clustering videos :
Ceph and a LXD cluster :
In a LXD cluster, by default, only the LXD database itself is internally replicated to the other cluster members. LXD doesn’t actually replicate any instance data, so that must be handled either by the application workloads or by the underlying storage driver. LXD supports Ceph storage driver for replicated, highly available storage.
Easy overlay networking in LXD with the Ubuntu Fan
LXD’s simplest overlay network for clustering is the bridge in fan mode:
OVN and a LXD cluster:
LXD also supports OVN overlay networking to allow for highly available, distributed virtual networks with support for ingress options:
Ingress proxy
Those are great thanks. I have watched many of them. I am stillfinding myself wanting a mote real world example. I have also noticed that most videos dont cover what happens when you need to migrate containers. They cover initial setup but not long term maintenance.
Thanks for your reply.
Maybe something that @stgraber can cover in future videos (cluster operations and ingress).
Do you have specific questions though?
To move an instance between cluster members you can do this:
lxc move <instance> --target=<cluster member>
Sure lets assume an ingress proxy is attached does that move when migrating? Or do i need to have some sort of process to change that?
I have been trying to understand a data volume connected to the container and how that migrated. I need to watch the ceph video. That might be covered there.
Then i have questions on backup and recovery of containers and the cluster itself. Container and local volume snaps are well covered. But what do you do to recreate/ recover if the original container was corrupted.
How do you use backups to recover a broken cluster?
Whats the most efficient way to copy ony changed snaps to an offsite location?
On the microceph announcement @stgraber mentioned he does a weekly rolling reboot for security updates. Does that involve container downtime? If not how was that achieved? What are the gotchas if any? What about container security updates? Is his process automated? A video or series talking about that cluster and its operations would be interesting.
Affitionally what firewall rules should be in place for cluster security? Can lxd proxy control the firewall?
There is a video on the ingress proxy here:
Yes it does move with the instance.
There is a section on instance backups here:
You can use lxc export and lxc import to backup/restore instances to tarballs.
Or you can use lxc copy --refresh to copy an instance (and its snapshots) incrementally to another LXD server, which you can schedule using an external scheduler (cron).
He just reboots everything.
Container security updates are expected to be handled by the system admin, and treated like any other machine in the system.
We have a video on production operation:
And a documentation section: