-
Title
After installing the Rapid Recovery agent on a Linux machine, the CPU utilization gradually increases to 100% -
Description
After installing the Rapid Recovery agent on a Linux machine, the CPU utilization gradually increases to 100%. We can see that this is caused by a mono process.
-
Cause
From the logs, Rapid Recovery agent restarted multiple times. We also found that the default port 8006 is in use by another application.
When rapidrecovery-agent starting, it starts listening to that port, but if the port is busy, starting of service can not finish successfully and watchdog send request to restart service. This is expected behavior. The starting of service is a rather complicated process for the system and its continuity affects the CPU. -
Resolution
The solution is to switch to another available port. Please refer to 185069 for instructions on how to configure the agent on Linux.
Mono is an open source framework that we use in order to make agent run in Linux, it’s part of our installation and it’s critical for our software operations. It’s like .Net for Windows. Lots of services can cause high CPU usage at the moment of service startup, our agent is one of such. We don’t have much control on how much of resources the system will allocate for the service. Spikes in resources consumption in general is expected behavior when something unusual happens to the processes\services\systems (like port conflicts in this case) both in Windows and Linux worlds.
A Change Request RR-109226 has been created to improve the logic for future releases. Possible change may include not trying to start the service indefinitely or restrict it to just a few retries.