Foglight Agent Manager 5.9.5 - Foglight Agent Manager Guide

Adding cartridges to the HA deployment whitelist

In addition to assigning an Agent Manager to an HA Partition, real HA activities do not take place until HA Aware agent types are deployed.

To provide eligibility for HA agent fail-over to existing agents that primarily function as remote monitors, the HA Deployment Whitelist property is included. This list-based property identifies prior cartridge deployments that were not originally marked as eligible for HA agent fail-over. In this list, each row entry contains a regular expression pattern that matches the agent package, version and agent types that are permitted to be managed in an HA Partition.

The criteria for inclusion into this list is that the agent type must be capable of monitoring remote resources. It does not make sense to enable HA fail-over for an agent that is collecting data that can only be retrieved from the Agent Manager host on which it is running. If such an agent is added to the Whitelist, when the host fails over, the agent can potentially start monitoring the resources of the newly promoted HA Primary, which can result in unexpected outcome when viewing the dashboards that display the collected metrics.

IMPORTANT: To effectively be eligible for fail-over functionality, agents must be using one of the RemoteConnection Services available in the Agent Manager Development Kit.

IMPORTANT: The elements in the expression must be separated with slashes ‘/’.

About agent fail-over

When the HA Primary host goes off-line, a schedule is registered that delays the transfer of the agents (and the promotion of another HA Primary) for two minutes. This delay is in place to allow the HA Primary host to be restarted and retain its HA status and assigned agents. If you find that the duration is too long, the following startup parameter can be set when launching the Management Server:

About non-HA deployments

Agent Managers that are assigned to an HA Partition can still manage deployments and agents that are not HA Aware. When these cartridges are deployed to an Agent Manager, they by-pass the HA Deployment Set checks and are only deployed to this host. If this host goes offline, any non-HA agents also become off-line.

Negotiating Agent Manager resources at runtime

Monitoring agents often have specific Agent Manager resource requirements. Some agents need more memory, others require larger queue sizes, or more lenient time-outs. The Agent Negotiation feature allows monitoring agents to request resource changes during deployment, activation, and runtime. During deployment, an agent package can request specific resource settings to applied as part of the deployment process. Depending on the current Agent Manager configuration, a restart may be required in order to apply the requested changes. This restart of the Agent Manager is automatic and is part of the deployment cycle.