Deploying Storage

How to Deploy

Enlist the chassis and deploy them for management and orchestration to install Swarm Storage. Swarm Platform supports any chassis equipped with IPMI 2.0 or newer.

1. Power on each chassis planned as a Swarm storage node. Each chassis PXE boots from the Platform server, enlists itself with Platform, and shuts itself down.

2. List the chassis currently available to run Swarm after enlistment completes:

   platform list nodes

   The number of chassis returned is the value to use in the next step for -n, unless some are kept in reserve.

3. Deploy Swarm Storage to the chassis being managed by Platform:

   platform deploy storage -v {#.#.# Version of Swarm Storage} -c {Path to cluster.cfg} -n {# Nodes To Deploy}

4. At this point, the Platform powers on each of the nodes twice. The first power on is commissioning, involving hardware interrogation and other tasks, which ends with a power off. The second power on is to boot Swarm Storage on the chassis.

5. Once Swarm is running, deploy Content Gateway, or, if not using Content Gateway, proceed to Deploying a Proxy.

   platform deploy proxy -b {Path to Service Proxy Bundle zip}

Power Control with ESXi 

It may be ideal to use local Storage VMs with a VMware ESXi hypervisor. These VMs enlist, but the power type is not detected by MAAS. Commissioning and deployment require the "Manual" power type (see next), and require manually powering on these VMs (shutdown works).

Using the native VMware power type is possible, but more complex: select it through the VMWare UI as well as the MAAS CLI, and use a valid (not self-signed) SSL certificate on the hypervisor. 

Using Manual Power Control

When a chassis is powered on for the first time, the Platform Server auto-detects which power mechanism to use when processing them for deployment. Switch the deployment to manual power control (add the --manual flag to the platform deploy command) and then manually power on machines at the right time as they go through the lifecycle if a chassis has IPMI for power control of the chassis but requires Platform server to ignore IPMI.

When using manual power control, use CLI commands to know when to power on each chassis, so use -y <system-id> flag, rather than -n <# of nodes>.

1. Run the deploy command for an individual chassis, using the  --manual and -y <system-id> flags.

2. Use the --state Commissioning flag to know when to turn the chassis on.

3. Power on the chassis as soon as the chassis appears in the list.

4. The chassis goes through commissioning after being powered on and then powers itself off.

5. Use the --state Deploying flag to know when to power on the chassis for the final time (to deploy the Storage software):

6. Power on the chassis for the final time as soon as the chassis appears in the list.

Changing Network Bonding Mode

The Platform server boots all Storage chassis using the "balance-alb" bonding mode by default. The default bonding mode used can be modified for future deployments by using the following command:

Use a mode for the Linux bonding driver that is tested and supported for Swarm:

Override the default value by using the --bondingMode flag to use a different bonding mode when deploying an individual chassis:

Run the add kernelparam command to modify the bonding mode for a chassis that's already deployed:

Restart the chassis after the command completes.

Configuring Subclusters

Assign chassis to subclusters after all chassis are deployed and are running.

Use the assign command to assign a chassis to a subcluster:

Add to subcluster

Use the list command to review the subcluster assignments:

List subclusters