Connectors integration steadily increases in memory consumption over time #3914
Description
Bug Description
In an agentless deployment, the connectors process memory usage increases after each full sync and does not return to baseline, eventually leading to an OOM kill.
To Reproduce
I would assume that the steps to reproduce are to deploy a Jira connector with an agentless agent.
But there has been an attempt to reproduce this by @artem-shelkovnikov without success. So it might be related to the Jira Cloud instance used.
Expected behavior
Connector process memory usage should return to a steady baseline after a full sync completes, without continuous growth across syncs leading to OOM.
Screenshots
Here is the memory usage of the entire agentless pod over a period of 7 days. Each bump happens when a full sync is triggered. The full sync on the pink arrow resulted in an OOM kill.
Here is the same period of time, showing the memory usage of each process in the pod:
Environment
- agentless container image:
docker.elastic.co/cloud-release/elastic-agent-service:9.2.3