19 Nov 2025 05:02 PM
Monitoring Applications. Dynatrace.
How many resources does OneAgent consume in full-stack mode on a Linux server?
I need to know how much RAM the Dynatrace agent consumes when applying full-stack monitoring on a Linux server. How much CPU resources can Dynatrace use? Can this affect the server's performance? Could it cause degradation on a Linux server?
Solved! Go to Solution.
19 Nov 2025 09:43 PM
Hi @VH
Have you read this page: https://docs.dynatrace.com/docs/ingest-from/dynatrace-oneagent/oa-requirements
If you need a more precise number you can maybe try to install OneAgent on an identical test server. Alternatively you can install it on the actual server in infra mode first and the tru to enable full-stack later to see the increase in memory usage.
I hope it answered your question.
20 Nov 2025 10:27 AM
Hi @VH
In full-stack mode, OneAgent is designed to run with a very small footprint and to stay below well-defined resource limits so it won’t interfere with the host’s own workloads.
Memory usage
On a Linux host, the combined memory consumption of all OneAgent modules (process monitoring, network, logs, deep monitoring, etc.) typically stays in the range of a few hundreds of megabytes at most. The exact value depends on the number of monitored processes, technologies, and traffic volume, but it remains stable and predictable.
CPU usage
CPU consumption usually stays below 3–5% even under load. OneAgent continuously adapts its sampling and analysis work based on host conditions, so it avoids competing with application processes.
Built-in protection mechanisms
OneAgent includes automatic safeguards: if the host’s CPU becomes constrained or if the agent detects that its own activity could impact the server, it will throttle or deactivate certain analysis components. This ensures that monitored workloads always remain prioritized and that Dynatrace does not degrade the machine’s performance.
23 Nov 2025 05:21 PM
The @JeanBlanc 's answer its really nice !
OneAgent has technically defined the resources required for different activities, but the final consumption depends on the various processes, technologies, and capabilities activated.
The best way to understand this is to install it in a real-time or time-controlled storage environment