Nodes Page#

To display the Nodes page, click Nodes > Compute Nodes in the left navigation panel or, from the Overview page, click the Manage Nodes link in the Nodes panel.

Nodes Page

At the top of the Nodes page is a Node Filtering subpanel (see Node Filtering). Below that is a summary of the total node count, the counts of "up" vs "down" nodes, and the counts of healthy vs. unhealthy nodes. These summary lines are clickable links that overwrite the node selector to select just the nodes in that summary. For example, clicking "7 unhealthy" selects just the nodes currently reported as "unhealthy".

If you are using a job scheduler, such as Slurm, you can toggle the Sched switch to show scheduler status. For details on installing job schedulers, see Job Schedulers. For details on configuring job schedulers to work with the ICE ClusterWare™ platform, see Monitoring Scheduler Info.

Below the Node Filtering subpanel is the Node Grid/Node List subpanel showing each node in a potentially filtered display. Each node is color-coded to indicate status and health to quickly highlight any anomalies.

The Create nodes text box is available at the bottom of the page. The only field required to add a new node to the cluster is the MAC address. Alternatively, you can supply a JSON string that adds multiple nodes.

For example, to create five compute nodes, you could provide:

[ {"mac":"FF:11:22:33:44:55"}, {"mac":"11:22:33:a4:fe:66"}, {"mac":"22:33:44:55:66:77"},
  {"mac":"00:99:88:77:66:55"}, {"mac":"ef:45:fd:43:23:54"} ]

Or you can create a node with multiple fields, such as:

{ "mac": "11:22:33:44:55:66", "attributes": { "_boot_config": "SlurmBoot" }}

Node Filtering#

Node Filtering allows you to view, create, and manage subsets of compute nodes and can be used to select subsets of the full set of nodes for display in Node Grid or Node List mode.

A filter can be defined and saved as a Dynamic Group, which is dynamic in the sense that a specific filtering criteria may show different nodes when applied at different times. You can also create a dynamic group from the Dynamic Groups Page.

The following actions are available in the Node Filtering section:

  • Select a dynamic group - Selecting an existing dynamic group from the Dynamic Group menu updates the list of displayed nodes per the Selector field.

  • Enter a selector expression - Once a valid selector is entered into the Selector field, the list of displayed nodes updates.

  • Select Manage Dynamic Groups from the Dynamic Group menu - Opens the Dynamic Groups page.

  • Select Clear from the Dynamic Group menu - Resets node filtering.

Node Grid Display#

Nodes Grid

The Node Grid is a row/column grid display where each node is represented by a cell. Above the grid display is a panel containing node status summary information and display options. Each node/cell conveys data with its visual properties:

  • Label - This is the node name that appears on the cell. Long names are truncated. The full name is visible in the Node cell.

  • Node status - The background color of the cell indicates node status: blue=UP, red=DOWN, orange=BOOTING, outline (no background color)=NEW.

  • Scheduler status indicator - This indicator is present only if Slurm is configured. Possible values: UNKNOWN, READY/IDLE, BUSY, UNUSABLE/ERROR. If the indicator is ERROR, there may be an accompanying error message. This is displayed when the user hovers over the Scheduler Status Indicator.

You can interact with cells in two ways:

  • Hovering: Exposes a pop-up containing additional identification and status information about the node.

  • Clicking: Exposes a pop-up containing more details than what is seen in the Hovering popup.

Node List Display#

Nodes List

This display is the unfiltered list of nodes. This example cluster has 49 total nodes. Note the "10 per page" in the lower right and the page number selectors in the lower left.

Above the table you can note the existence "7 unhealthy" nodes. Suppose we filter for those nodes:

Nodes List Unhealthy

Now you see just the seven unhealthy nodes, and in their Attributes you can see why the health checking script has decided these are unhealthy: "Scratch space is not mounted".

Immediately above the table showing each node (as filtered or unfiltered) is an "Actions" text box. When the admin selects one or more nodes (either individual nodes by clicking on the box to the left of the node name, or all nodes by clicking on the box to the left of the title "Name"), then clicking in the "Select action" text box exposes several possible actions that can be taken against all the selected nodes. To the right of the "Actions" select menu indicates precisely how many nodes will be subject to the selected action.

The available actions:

  • Select action

  • Execute…

  • Soft Control: Reboot, Reboot - kexec, Reset, Shutdown

  • Hard Control: Power On, Power Off, Cycle Power

  • Other: Delete