Saltar al contenido principal

Start and stop stores

After you create a store, you must start it before any client can connect, and you stop it when it is no longer needed. You can do this from PowerShell or the Web Console; either way the workspace's managers launch or shut down the node processes.

Activate a workspace first (see Connect to a cluster).

Store lifecycle​

A store moves through four states:

  • Created — the configuration and node topology exist, but no process is running.
  • Started — the manager has launched the nodes.
  • Running — the nodes have formed the cluster and accept client requests.
  • Stopped — the nodes are shut down.

You can move a stopped store back to running at any time by starting it again.

Start a store​

Start-ZrStore starts the nodes of a store. With only -Name, it starts every node; add -InstanceName to start a single node.

Start-ZrStore -Name <StoreName> [-InstanceName <NodeName>]

To start all nodes of a store:

Start-ZrStore -Name orders-cache

Internally the cmdlet queries each manager for the store's nodes, then sends each node a start command. A successful run prints one row per node, with the action labeled StartInstance:<nodeName>:

Manager                Action                         Result  Message
---------------------------------------------------------------------
10.0.0.11:7801 StartInstance:10.0.0.11-n0 SUCCESS Started
10.0.0.12:7801 StartInstance:10.0.0.12-n0 SUCCESS Started

A healthy result shows SUCCESS for every node. Confirm the store is fully up with Get-ZrStore — a healthy store reports status Running with all instances running. See View store information.

To start just one node — for example, to bring back a single node you stopped:

Start-ZrStore -Name TestStore -InstanceName 10.0.0.11-n0

Stop a store​

Stop-ZrStore shuts down the nodes of a store. Like Start-ZrStore, it stops all nodes by default, or one named node with -InstanceName.

Stop-ZrStore -Name <StoreName> [-InstanceName <NodeName>] [-Graceful]

To stop all nodes:

Stop-ZrStore -Name orders-cache

To stop a single node:

Stop-ZrStore -Name orders-cache -InstanceName 10.0.0.11-n0

Because a whole-store stop takes every node down together, there is no surviving peer to hand off to, so Stop-ZrStore defaults to a fast hard kill. Add -Graceful to opt into a per-node drain and cluster-leave instead — useful when you are stopping only one node of a still-running store:

Stop-ZrStore -Name TestStore -InstanceName 10.0.0.11-n0 -Graceful

Stopping a store disconnects all connected clients. Store data is held in memory, so stopping the store clears it unless it is persisted externally.

nota

To stop a single node with a graceful drain as the default behavior, use Stop-ZrNode -StoreName <store> -Name <node> — a single-node stop drains and leaves the cluster by default, and its -Force (alias -HardKill) switch hard-kills instead.

Execution options for large clusters​

Start-ZrStore and Stop-ZrStore accept two options that help on multi-server clusters:

  • -Parallel sends the operation to all managers concurrently instead of one at a time.
  • -FailFast stops the whole operation the moment any node reports a failure, rather than continuing through the rest.

Troubleshoot start and stop​

SymptomCauseFix
"No managers connected" errorNo workspace is activeActivate a workspace, or pass -Managers
Store does not start; a node failsIts cluster or client port is already in useFree the port or re-create the store on unused ports; see Ports and networking
Store starts but clients cannot connectThe store is not fully running, or the client port is blockedVerify with Get-ZrStore; open the client port range in the firewall
A node must be taken down cleanlyYou want a drain and handoff, not a hard killStop the single node with Stop-ZrStore -InstanceName <node> -Graceful, or Stop-ZrNode

Next steps​