Permit

Hey <@U0450KJGEQ7> :wave: Someone from our team will be with you shortly! :relaxed:

Hi <@U0450KJGEQ7>, the architecture supports any number of opal server and opal clients running together.
It is good to have at least 2 in each so you will have some mature of redundancy, it is also a factor of the amount of changes in the policy and the size of it, and the size of the server.

If you can share more details about your setup we can try to help a bit more.

I want to know how scaling should work with opal. If I need to scale to 100 opal clients, I would assume I would have more than 1 opal server. What kind of metrics should I use? 1 server per 10 clients?

I know that there can be other factors such as cpu, memory and network, just want a good rule of thumb to go off of.

First of all one OPAL server can technically support 100s if not 1000s of OPAL clients. I would stay in the 100s at most just to be safe. We are aware of some companies who are using OPAL in production that decided on a 1:10 ration (unnecessary IMO but it is something i've seen) - so for example a real case was 9 OPAL servers and about 95 clients.

Awesome great to know. Sounds like that ratio is pretty high then. I was afraid of potential networking bottlenecks. Should I still have at least 2 servers in case one goes down? Then transfer all of the clients to the online server?

if you run more than one opal server you need to connect them through a broadcast backbone service (i.e: postgres, redis) and expose all opal servers behind a load balancer.

the opal client needs to get the opal server address hard-coded so you need a load balancer