An integration is returning thousands of metrics, or is running a large number of check instances. You can see a summary of the running check instances, as well as the number of metrics collected, by running the status CLI command and checking the Collector section.
The Agent’s Python or Go runtime is causing high resource consumption. Enable Live Processes Monitoring to check if the Agent process is consuming unexpected amounts of memory or CPU. You can also use your operating system’s activity manager to check Agent process resource consumption.
The Agent’s behavior is triggering Windows anti-malware or antivirus tools, causing high CPU usage.
The Agent is forwarding a very large number of log lines or DogStatsD metrics.
Adjustments to reduce resource usage
Here are some adjustments you can make to your Agent configuration to reduce resource usage:
For integrations that have many check instances or are collecting large numbers of metrics, adjust the min_collection_interval in the integration’s conf.yaml file. In general, the Agent runs each check instance every 10 to 15 seconds. Setting min_collection_interval to 60 seconds or more can help reduce resource consumption. For more information on the check collection interval, see the Custom Agent Check documentation.
Check if an integration is configured to use Autodiscovery, or if an integration is using a wildcard (*) that could be scoped more specifically. For more information on Autodiscovery, see Basic Agent Autodiscovery.
Reach out to Datadog Support
If none of the above solutions are right for your situation, reach out to Datadog Support. Make sure you’ve enabled Live Processes Monitoring to confirm that the Agent process is consuming unexpected amounts of memory or CPU.
When opening a ticket, include information on how you are confirming the issue and what steps you have taken so far. Depending on whether or not you can isolate the problem to a single integration, include information from one of the following sections.
High consumption isolated to a single integration
If only one integration is consuming high amounts of memory, send a debug-level flare along with Python memory profile output:
High consumption not associated with a single integration
If the high memory consumption is not associated with a single integration, send a debug-level flare with a profile, collected during a period when the Agent is using more memory or CPU than expected:
