Ports and networking
Clustron Zaris uses three kinds of ports: one for management, one for node-to-node communication, and one for client connections. Knowing which port carries which traffic is essential for opening firewalls correctly and for diagnosing why nodes will not join a cluster or why clients cannot connect.
The three port ranges
Zaris divides the 7801–7899 span into three contiguous ranges, one per kind of traffic. Keeping them separate means you can reason about — and firewall — each traffic path independently.
| Range | Kind | Purpose | Direction |
|---|---|---|---|
7801–7810 | Management / ops | Administrative commands, the manager-to-manager control plane, the Web Console (7810), and each node's health probes (7802, /livez + /readyz) | Inbound to each manager (admins, other managers, browsers); health probes are local to each node |
7811–7859 | Cluster | Node-to-node coordination and data movement | Between nodes |
7861–7899 | Client | Application and ClientShell connections | Inbound to each node, from clients |
Each range is sized to allow many nodes on one server without collisions:
- Management / ops (
7801–7810). The manager's management API is7801; the Web Console sits at7810inside the same range and serves over HTTP by default, switching to HTTPS only when you enable control-plane HTTPS withEnable-ZrManagementHttps. Each node also serves its health probes at7802(/livez+/readyz) — an operational endpoint consumed by the orchestrator (for example the Kubernetes kubelet), not client or cluster traffic. - Cluster (
7811–7859). Node-to-node ports begin at the-BaseClusterPortyou set onNew-ZrStore(default7811) and increment per node. - Client (
7861–7899). Client ports begin at-BaseClientPort(default7861) and increment per node.
The single-port gap (7860) between the cluster and client ranges is reserved headroom and is not used by default.
How the ports work together
Each traffic path uses exactly one of the port kinds:
- An administrator or the Web Console reaches a manager on the management port (
7801/7810). - A node reaches its peer nodes on their cluster ports.
- A client application reaches a node on that node's client port.
Admin / Console ── management port (7801/7810) ──▶ Manager
Client app ── client port (7861+) ──▶ Node
Node ◀─ cluster port (7811+) ──▶ Node
Firewall configuration
The installer automatically opens the full range that Zaris uses for management, cluster, and client traffic:
7801 – 7899
It opens this range on the local machine only. That is enough for a single-server cluster, where all traffic stays on one host and no external networking is involved — just make sure no other application already holds the ports.
If you plan to run a distributed cluster across machines, make sure these ports are reachable between them. Node-to-node and client traffic cross the network, so any host firewall, network firewall, or cloud security group between the servers (and between clients and the servers) must permit:
- the management range (
7801–7810) between administrators and every manager, - the cluster range (
7811–7859) between the nodes, and - the client range (
7861–7899) from clients to the nodes.
Use reachable IP addresses or resolvable hostnames for multi-server clusters — never localhost, which only refers to the local machine.
On multi-server clusters, the installer does not open ports on the network firewalls or cloud security groups between machines. If node-to-node traffic on the cluster range is blocked, nodes fail to join and the cluster never forms. Open the cluster and client ranges across the network before you start the store.
Troubleshoot port and networking issues
| Symptom | Cause | Fix |
|---|---|---|
| Store fails to start; a node will not launch | A chosen port is already in use, or ranges overlap | Pick free -BaseClusterPort / -BaseClientPort ranges; avoid overlapping stores |
| Clients cannot connect | The store is not running, or the client port is blocked | Confirm the store is Running; open the client range; use the correct <host>:<clientPort> |
| Nodes do not join the cluster | Cluster ports blocked between machines | Open the cluster range across the network; verify connectivity between servers |
| Cannot connect to a manager | Management port 7801 blocked | Allow inbound 7801 on the server and any firewall in between |
Next steps
- Monitoring and metrics — confirm traffic is flowing once ports are open.
- Troubleshooting — full symptom-to-fix reference.
- Multi-node configuration — how ports map to nodes and servers.