View the Status of a Cluster at a Glance

See the general status of HPC topics/nodes to understand what your HPC system is doing.

View the Status of a Cluster's Nodes

View a graphical representation of the status of a cluster's nodes.

  1. Click the Monitor tab.
  2. Click the Nodes tab from the PBS Professional menu located on the left-hand side of the web page.
    A summary of all HPC clusters are displayed. Each cluster is represented by a row containing the cluster name, the total number of nodes, the number of queues within the cluster, and a colored bar graphically depicting the status of the nodes.
    Tip: By default, the latest job and node data is automatically retrieved and displayed every 5 minutes. Refresh the data manually by clicking .

    Refresh Data
    Figure 1. Refresh Data
  3. Optional: If the cluster's queues are not visible, click the name of the cluster to expand it.
    The queues attached to the cluster nodes are displayed. Each queue is represented by a row containing the queue name, the number of nodes attached to the queue, and a colored bar graphically depicting the status of the nodes attached to the queue.

Understand the Node View

A high-level view of the status of an HPC cluster's nodes.

A key step in monitoring and maintaining cluster health is to identify any deviance from normal operational state or performance. The Node View provides a view to the status of a cluster or a queue at a glance so that you can identify problem nodes, and drill down to node details for further investigation. A colored bar graphically displays the condition of the cluster or queue.

Each cluster is represented by a row containing the cluster name, the number of nodes and queues within the cluster, and a colored bar graphically depicting the condition of the cluster's nodes. The length of a color indicates the ratio of nodes in each condition.

Color Condition State
Red requires attention down, stale, state-unknown, unresolvable
Green in use job-busy, job-excl, resv-exclusive
Purple unavailable offline, provisioning, wait-provisioning
Yellow available free
Orange idle idle

Node view
Note: indicates the credentials of the user that was used when adding the cluster have changed server-side, therefore the cluster is registered but unreachable. indicates that there are not enough licenses available to show cluster summary or detail information.
Figure 2. Node View

You can drill-down to get detailed node information about all the nodes in a cluster, nodes attached to queue, or a subset of nodes in either of these categories filtered by condition.