Workspace
A workspace is a client-side named connection to a set of managers. It is how the CLI and web console target a cluster; it is not a server and not a store.
What a workspace is
A workspace is a named pointer, saved on the machine you administer from, that records which managers your admin commands talk to. It holds a name, the list of manager endpoints (each written as host:port, defaulting to management port 7801), an optional admin token, and the cluster's deployment mode. You create one with New-ZrWorkspace and make it the active target with Use-ZrWorkspace.
A workspace is purely a client-side bookmark. Creating or switching a workspace does not start, stop, or change any server, node, or store — it only changes which managers your session addresses. The PowerShell workspace is the peer of the web console's workspace: both hold the same shape (id, name, managers, token) so the two tools describe the same cluster the same way.
# Create a workspace named "prod" targeting two managers, then activate it
New-ZrWorkspace -Name "prod" -Managers "10.0.0.10:7801","10.0.0.11:7801"
Use-ZrWorkspace prod
What a workspace is not
Because the terms are easy to confuse, it helps to state the boundaries directly.
- A workspace is not a server. A server is a real machine running Zaris; a workspace is a list of the servers' management endpoints, stored on your side.
- A workspace is not a store. A store is the data you read and write. One workspace targets a cluster that can host many stores.
- A workspace is not a cluster. A cluster is the set of running nodes. The workspace just names how to reach the managers that observe or supervise that cluster.
Creating versus joining
New-ZrWorkspace and Use-ZrWorkspace do different things, and the distinction matters.
New-ZrWorkspace creates a workspace by claiming currently-unclaimed managers into it. It rejects managers that already belong to another workspace, so it cannot silently take over an existing cluster.
Use-ZrWorkspace activates a workspace. If the named workspace is already saved locally, it just switches to it with no network call. If it is not saved — or you pass a seed manager address — it joins the existing workspace: it reads that manager's binding, verifies the name, discovers the full manager roster, saves it, and activates. Use-ZrWorkspace never creates or claims.
# Join an existing "prod" workspace from one of its managers, then activate it
Use-ZrWorkspace prod @('10.0.0.11:7801')
Deployment mode
A workspace records the deployment model of the cluster its managers front: supervisor or attach. In supervisor mode each manager forks and owns the node processes on its own machine, and the console and CLI aggregate the managers' views into one picture. In attach mode the nodes run under an orchestrator and the manager only observes. When you do not state the mode explicitly, New-ZrWorkspace auto-detects it by probing the first manager, and all of a workspace's managers must report the same mode.
How the terms relate
A workspace sits at the top of the addressing chain: it names the managers, the managers front a cluster, the cluster runs one or more stores, and each store's data is split into partitions and replicated across nodes. The table contrasts all seven building blocks.
| Term | What it is | Identified by | Where it lives |
|---|---|---|---|
| Workspace | A client-side named connection to a set of managers | Name + manager list | On your admin machine (~/.clustron) and the console |
| Server | A machine running the Zaris management service | Management address, host:7801 | One per machine (supervisor mode) |
| Store | The logical unit you read and write against | Store name | Presented across the cluster |
| Cluster | The set of nodes that together run one store | Its member nodes | Across one or more servers |
| Node | A single running store process | <host>-n<index> | One per partition-replica slot |
| Partition | A slice of the keyspace | A code word (Amber, Onyx, …) | Owned by one node at a time |
| Replica | A copy of a partition held on another node | Its partition + holding node | On a node other than the primary |
A workspace should point at managers that all observe the same cluster. New-ZrWorkspace warns if a manager you add already belongs to another workspace, because a manager is meant to front exactly one cluster.