A look into IBM Spectrum Symphony multicluster

The multicluster feature in IBM® Spectrum Symphony Advanced Edition allows you to connect multiple IBM Spectrum Symphony clusters into a federation cluster. Using this feature, you can leverage workload placement, to redirect IBM Spectrum Symphony sessions to non-busy clusters, balancing workload so that tasks within one session can be dispatched to multiple clusters based on resource availability on each federation member cluster. You can also deploy and manage service packages from the multicluster management console across member clusters. Finally, you can monitor workload and resources, as this feature collects data from IBM Spectrum Symphony clusters and sends it to the IBM Spectrum Symphony multicluster primary cluster for aggregation and display.

This feature is is available for clusters running IBM Spectrum Symphony Advanced Edition for IBM Spectrum Symphony workload (MapReduce applications are not supported).

The clusters that you want to join the IBM Spectrum Symphony multicluster must have hosts running the same version of IBM Spectrum Symphony Advanced Edition (for example, all hosts within multicluster can use IBM Spectrum Symphony 7.3.2). IBM Spectrum Symphony clusters running the IBM Spectrum Symphony Developer Edition, or Standard edition cannot join multicluster.

For IBM Spectrum Symphony multicluster installation steps and details, refer to Installing IBM Spectrum Symphony multicluster.

The following diagram depicts an IBM Spectrum Symphony multicluster environment:
Platform Symphony MultiCluster environment

Multicluster primary cluster

The multicluster primary cluster hosts the multicluster primary services and GUI. It contains a primary management host running IBM Spectrum Symphony. This primary management cluster oversees the federation cluster (made up of member clusters) to manage multicluster features such as workload placement and service package deployment across member clusters. The primary cluster resides on a set of hosts to ensure failover, and uses EGO technology for high availability and for managing services within multicluster. It is typically a one- or two-host cluster that runs four EGO services: the IBM Spectrum Symphony multicluster primary daemon, the multicluster management console, the EGO REST server, and the repository service (RS).

There is only one primary host in the multicluster primary cluster. The primary candidate is a host that can become the primary host should the multicluster primary host fail. If the primary host fails, one of the primary candidate hosts automatically takes over the primary host role.
Tip: Refer to Recovery for a failed multicluster primary cluster for details about creating a backup multicluster primary cluster in case the multicluster primary cluster fails.
The multicluster primary cluster is composed of the following components:
Multicluster primary
The multicluster primary communicates with the member clusters using the multicluster proxy service on each member cluster. For example, the multicluster primary cluster gathers resource information collected by each multicluster proxy service on each member cluster. The multicluster primary cluster can then accept resource repurposing requests, and invokes the multicluster proxies to move the resources. It also provides a cluster registration service for adding clusters to multicluster or removing clusters from multicluster.
VEMKD
The EGO kernel daemon (VEMKD) daemon supports host failover and is responsible for starting EGO services including the SMCP service (for the multicluster proxy service on member clusters) and the WEBGUI service (for running the multicluster management console).
Load Information Manager (LIM)
The Load Information Manager (LIM) daemon starts VEMKD on the multicluster primary host.
Multicluster management console
The multicluster management console can be accessed through a web browser. The minimum screen resolution required is 1024 x 768. The web server runs as the WEBGUI daemon. Use the multicluster management console to manage all your multicluster tasks; the multicluster management console header includes links to the three main areas:
Workload
Under the Workload menu, a multicluster user can work with the applications they have been authorized to manage, including these tasks:
  • Access an overall dashboard view of the applications within multicluster.
  • View a list of applications within multicluster.
  • Manage the settings for applications and workload bins. For application workload placement, you can place applications into workload bins; each bin can have rules and filters that define the workload placement for the application.
Resources
Under the Resources menu, a multicluster administrator (or a user with the SMC_CLUSTER_CONTROL permission) can manage clusters, including these tasks:
  • Access an overall dashboard view of the clusters within the multicluster.
  • Manage the clusters that are current members of the federation cluster (a group of IBM Spectrum Symphony clusters managed by multicluster), or clusters waiting to participate.
  • Manage resource plans, including repurposing hosts to move hosts between clusters and monitor progress.
  • Manage a global resource view so that you have a single point of access to all the resource plans in all your clusters monitored in multicluster. You can view planned allocations per time interval for all consumers.
System
Under the System menu, the multicluster administrator (or a user with the SMC_CLUSTER_CONTROL permission) manages multicluster settings, including these tasks:
  • Manage multicluster operations.
  • Manage workload placement settings, including enabling clusters and applications to participate in, and defining global rules for, workload placement.
  • Manage dashboard display settings.
  • Manage users, including creating multicluster accounts and assigning multicluster user roles.

Additionally, to obtain help information on using any of the multicluster management console pages, click the question mark icon (Help (question mark) icon to launch the Knowledge Center with the MultiCluster Management Console) on the header and click Documentation to open IBM Documentation.

Multicluster member clusters and the multicluster proxy service

You can have one multicluster primary cluster, but one or more multicluster member clusters (in the diagram, there are two: Silo Cluster A and Silo Cluster B). For each cluster that you want to manage, you enable a multicluster (SMCP) proxy service on a management host; this services is responsible for discovering hosts in the cluster and triggering actions to repurpose them. There is one multicluster proxy service per cluster.

Technically, both the multicluster primary cluster and the multicluster repository service (RS) need to connect to the multicluster proxy service on the member clusters. For simplicity, multicluster connections refer to the connections between the multicluster primary cluster and the member clusters. For more information about the multicluster proxy service and connections see Multicluster proxy service and Multicluster connections.

The multicluster agent is a daemon that runs on demand on a host being repurposed. It is responsible for stopping and starting LIM, and starting move-in and move-out scripts and reporting their results to the multicluster proxy.

The multicluster primary cluster periodically checks for disabled clusters and removes them from the database if they do not connect to the multicluster primary cluster within 120 seconds. Disabled clusters are removed from the cluster view in the multicluster management console.
Note: Multicluster does not support the use of duplicate host names across different clusters.