When basic agent metrics such as Java Version are missing, along with whole categories such as JMX or PMI, this often indicates that an agent clamp has been triggered. The example shows metrics that will commonly disappear when a clamp is triggered. On the left is an agent as it appears normally. On the right, after the clamp. Circled categories on the left highlight missing metrics after a clamp is triggered.
Metric clamps can be triggered on both the agent or EM side, but when basic metrics such as Java Version are missing, this usually indicates a clamp on the EM side. The APM Status Console screen (if authorized for your account) can show an EM administrator which collectors can possibly have clamped agents. To see this, click on a collector icon to see a list of clamps (if any) are triggered.
Once the collector that the agent is connected to is determined, the problem can be temporarily solved by raising the metric limits for the collector. To do this, find the file /<EM_HOME>/config/apm-events-thresholds-config.xml. Next, make a backup copy so any changes can be rolled back easily. Then, open the file in a text or xml editor, and locate the following two clamp settings:
These values control the per agent limit (...agent.metics.limit) and the limit for all agents in aggregate (...metrics.live.limit).
Doubling them as a temporary measure may allow the agent to appear fully in the investigator tree. To change the limit, locate the line below the clamp id called a threshold value. For example, change
These are hot properties, so once the value is changed, and the file updated (save the file), the property change takes effect in a few minutes.
Once the agent appears fully, look at the metric counts to determine where the bulk of the metrics are coming from (sql, jmx, pmi, etc). The agent should be configured to reduce those types of metric (through sql normalization/configuration, filtering out unneeded jmx or pmi metrics, etc).
If an EM administrator decides to raise the agent metric threshold values on a permanent basis, caution should be exercised as the increased metric volume can cause EM performance problems. Monitor the EM performance carefully, and reduce thresholds if needed.