Can bitwarden_rs tolerate being clustered?

[nneul]

Level 1:

Will it’s implementation tolerate having two or more instances running against same database (mariadb) for failover purposes, with only a single instance actively serving users (though both running) at same time?

Level 2:

Are it’s internal operations of the server stateless outside of backend database, such that it could tolerate being hit from multiple container instances at the same time? (Assume round-robin or without-warning failover between nodes.)

In both cases, key question is whether any in-memory state is maintained that could be corrupted by having two containers running at same time against same DB?

[nneul]

Found this, so looks like that answers my question:

Is the bitwarden rs service stateless if its using postgres backend?

I’m deploying on Kubernetes. I was hoping to deploy the Bitwarden server in a deployment with 3 replicas and then have a single stateful state postgres instance with an open ebs volume. Do bitwarden rs instance have any state to worry about if the postgres db is not colocated? One reason I ask is because the docs point at this setup as an example: https://github.com/icicimov/kubernetes-bitwarden_rs/blob/7b6c0d49f9c5f5c6b7fce31a2a97bca9df8d0d25/statefulset.yml#L2-L3 p.s … Just setting the postg…

[BlackDex]

My answer on that post you refer to still applies.
So, in short, yes it could work with some caveats to keep in mind.

• Attachments probably need to be on some shared storage. (Same goes for icons etc…)
• The rsa key’s used for the JWT need to be shared.
• Websockets will not work fully if using round-robin.