Search the docs

Shut Down and Power Up a Cluster

Applies to: Hammerspace 5.2 and later.

You may need to take a whole cluster offline to upgrade the underlying hardware, or to do maintenance on a virtual machine host.

Stop all client access before shutting the cluster down. These steps take every cluster service offline.

Bringing a cluster that participates in a Global File System back online means catching its metadata database up from the other participants. Wait at least 30 minutes after the other participants are online before relying on it.

Contact Hammerspace Support if you have questions about taking nodes offline.

Shut Down the Cluster

Shutting down takes two steps: one command inside the cluster, and one power-off outside it.

  1. Connect to the cluster over SSH with an administrative account.

  2. Shut the cluster down.

    Command:

    system-shutdown --force

    The DSX nodes shut down one after another, then the secondary Anvil. The primary Anvil is left in maintenance mode rather than powered off.

  3. Power off the primary Anvil. It is the only node you power off from outside the Hammerspace admin shell.

    Command:

    shutdown -P now

    On a hypervisor, powering the virtual machine off from that interface is equally safe.

Power the Cluster Back Up

  1. Power on one Anvil node.

  2. Wait 30 seconds, then power on the second Anvil.

  3. Once both Anvil nodes are online, power on the DSX nodes.

Only the sequence matters. Use whichever method suits the environment — the power button, an IPMI or BMC interface, or the hypervisor console.
Run any CLI command with --help for its full syntax.