I want to deploy a docker swarm network with 2 organizations, 2 peers each and a CA. All of the nodes will be connected and communicate via the docker swarm network. To my understanding, each organization would have a manager with worker and the workers will host the peer and CA containers.
But, it seems to me that a leader manager can administer the whole swarm. That would mean that an organization manager could access another org's workers. This shouldn't be the case, so what am I missing?
If you run nodes for multiple organizations on a single Swarm network (or Kube cluster or VMs in the same cloud account, etc), then of course whoever has access rights to the deployment can gain access to the nodes of all organizations.
This might be ok if a single entity is hosting all the nodes for a given blockchain network (although it does potentially defeat some of the tenets of blockchain) as typically individual orgs will only have access to their blockchain nodes via the Fabric APIs and won't have system-level access.
Typically, deploying multiple orgs to a single cluster is done for development, testing or experimental purposes. In a real-world deployment, each organization is responsible for its own compute infrastructure.