Book Image

Mastering vRealize Operations Manager

By : Chris Slater, Scott Norris
Book Image

Mastering vRealize Operations Manager

By: Chris Slater, Scott Norris

Overview of this book

Table of Contents (23 chapters)
Mastering vRealize Operations Manager
About the Authors
About the Reviewers
Just Messing Around

vRealize Operations Manager node types

vROps 6.0 contains a common node architecture. Before a node can be added to an existing cluster, a role must be selected. Although deployment will be discussed in detail in the next chapter, from a design perspective, it is important to understand the different roles and which deployment model best fits your own environment.

The master / master replica node

The master or master replica node is critical to the availability of the Operations Manager cluster. It contains all vROps 6.0 services, including UI, Controller, Analytics, Collector, and Persistence, as well as the critical services that cannot be replicated across all cluster nodes. These include:

  • Global xDB

  • An NTP server

  • The GemFire locator

As previously discussed, Global xDB contains all of the data that we are unable to shard across the cluster. This data is critical to the successful operation of the cluster and is only located on the master node. If HA is enabled, this DB is replicated to the master replica; however, the DB is only available as read/write on the master node.

During a failure event of the master node, the master replica DB is promoted to a full read/write master. Although the process of the replica DB's promotion can be done online, the migration of the master role during a failover does require an automated restart of the cluster. As a result, even though it is an automated process, the failure of the master node will result in a temporary outage of the Operations Manager cluster until all nodes have been restarted against the new master.

The master also has the responsibility of running both an NTP server and client. On initial configuration of the first vROps node, you are prompted to add an external NTP source for time synchronization. The master node then keeps time of this source and runs its own NTP server for all data and collector nodes to sync from. This ensures that all the nodes have the correct time and only the master/master replica requires access to the external time source.

The final component that is unique to the master role is the GemFire locator. The GemFire locator is a process that tells the starting or connecting data nodes where the currently running cluster members are located. This process also provides load balancing of queries that are passed to data nodes that then become data coordinators for that particular query. The structure of the master node is shown in the following figure:

The data node

The data node is the standard vROps 6.0 node type and is the default when adding a new node into an existing cluster. It provides the core functionality of collecting and processing data and data queries as well as extending the vROps cluster by being a member of the GemFire Federation, which in turn provides the horizontal scaling of the platform.

As shown in the following diagram, a data node is almost identical to a master/master replica node with the exception of Global xDB, the NTP server, and GemFire locator:

The remote collector node

The remote collector role is a continuation of the vCOps 5.x Installable concept, which includes using a standalone collector for remote sites or secure enclaves. Remote collectors do not process data themselves; instead, they simply forward metric data to data nodes for analytics processing.

Remote collector nodes do not run several of the core vROps components including:

  • The Product UI

  • Controller

  • Analytics

  • Persistence

As a result of not running these components, remote collectors are not members of the GemFire Federation, and although they do not add resources to the cluster, they themselves require far fewer resources to run, which is ideal in smaller remote office locations.


An important point to note is that the adapter instances will fail over other data nodes when the hosting node fails even if HA is not enabled. An exception to this is the remote collectors, as adapter instances registered to remote collectors will not automatically fail over.