a fabric should have a UUID so that if fabric is run on some infrastructure (e.g. kubernetes) we can filter all containers for a specific fabric instance easily

[jstrachan]

it'd be nice to have a unique fabric ID for a fabric; which we could then tag containers so that any general purpose docker / kubernetes style IaaS can then easily query the related containers for a specific fabric ID.

We could also use the UUID in the ZK / etcd path so that you could run several logical fabrics on the same underlying ZK / etcd cluster while keeping them separate.

e.g. rather than using /fabric/containers/ we'd use /fabric/$UUID/containers.

This will also help the multi-data-centre support (#622); since each data centre fabric would be logically separated; then we can easily mirror remote data centre fabric data into local ZK / etcd in read only mode etc.

[davsclaus]

Is this relevant anymore as on v2 we use kube namespaces?