High availability with Unraid for certain docker containers

[Igor Papirov]

I have a number of servers in my homelab all running Unraid.

There are a number of containers in my setup that I'd like to ensure are always working, even if one of the servers fails.  Specifically these are:

• Home Assistant
• Cloudflared
• Nginx proxy manager
• Overseer
• Plex
• and a few more auxiliary things

 

What's the best way to accomplish this? Especially, since certain containers should not have more than one instance running (like Home Assistant) From what I've been reading, I should probably setup VMs with keepalived and dockerswarm, but is there a better way to do this in Unraid natively?  I would /really/ prefer the latter option

Also, I think I need to have a global NFS share to serve as shared storage for these containers.. how do I accomplish an HA mode for that?

 

TIA

Edited February 6, 2023 by Igor Papirov

[apandey]

HA is very subjective, depending on what failures you are willing to accept

Since these are stateful workloads, let's start with HA on storage (shared nfs) - if you want to protect against individual disk failures, any raid or unraid setup is enough. If you want protection against controllers or servers themselves, you will need to look at a distributed storage solution. Then again, are they in same rack or share power distribution etc is something you need to further decide. At this point, it is already a problem beyond unraid which is limited to a single server. Mostly, you will need to look at individual application and decide based on its HA architecture too

 

You also mention running one instance at a time, so sounds you are looking for automated failover rather than HA. In that case, you should also consider how much switching downtime you can tolerate. If it's not too tight, sometime it's cheaper to reinstall from a backup than to have a full HA solution

[Igor Papirov]

4 hours ago, apandey said:

HA is very subjective, depending on what failures you are willing to accept

Since these are stateful workloads, let's start with HA on storage (shared nfs) - if you want to protect against individual disk failures, any raid or unraid setup is enough. If you want protection against controllers or servers themselves, you will need to look at a distributed storage solution. Then again, are they in same rack or share power distribution etc is something you need to further decide. At this point, it is already a problem beyond unraid which is limited to a single server. Mostly, you will need to look at individual application and decide based on its HA architecture too

 

You also mention running one instance at a time, so sounds you are looking for automated failover rather than HA. In that case, you should also consider how much switching downtime you can tolerate. If it's not too tight, sometime it's cheaper to reinstall from a backup than to have a full HA solution

 

Everything is installed in one rack. I am looking to protect against container failures and against server failures.  If the power goes out and my UPSs run out, then it is what it is.  However, if one of the server crashes (as they sometimes do), I'd like for critical containers to continue to function.

[jmagar]

OP have you made any progress on this at all?? I'd like a setup pretty much exactly like you want. I've got my unRAID server and a couple PIs that I'd like to cluster together somehow either via docker swarm or kubernetes or whatever and I'd like to run nextcloud, vaultwarden, freshrss, swag and a couple other containers and have it so I can still use them if my unRAID server has to go down. 

[rbrowning85]

Any progress on this topic?? I am looking to do the something similar as well. 

[SimonF]

50 minutes ago, rbrowning85 said:

Any progress on this topic?? I am looking to do the something similar as well.