An implementation of a distributed access-control server that is based on Google Zanzibar - "Google's Consistent, Global Authorization System".
An instance of an access-controller is similar to the aclserver implementation called out in the paper. A cluster of access-controllers implement the functional equivalent of the Zanzibar aclserver cluster.
An access-controller server supports single node or multi-node (clustered) topologies. Instructions for running the server with these topologies are outlined below.
To gain the benefits of the distributed query model that the access-controller implements, it is recommend to run a large cluster. Doing so will help distribute query load across more nodes within the cluster. The underlying cluster membership list is based on Hashicorp's memberlist
a library that manages cluster membership and member failure detection using a gossip based protocol.
A cluster should be able to suport hundreds of nodes. If you find otherwise, please submit an issue.
$ ./access-controllerStart a multi-node cluster by starting multiple independent servers and use the --join flag
to join the node to an existing cluster.
$ ./access-controller --node-port 7946 --grpc-port 50052
$ ./access-controller --node-port 7947 --grpc-port 50053 --join 127.0.0.1:7946
$ ./access-controller --node-port 7948 --grpc-port 50054 --join 127.0.0.1:7947A Helm chart is included in this repository to provision an access-controller cluster in Kubernetes.
helm install access-controller ./helm/access-controller