On instances running dr_agent for FoundationDB 7.3.37, we’ve observed that over the course of 5 months dr_agent will very slowly leak memory (around 50MB a day). We did not observe any similar behaviour on other nodes that were not running dr_agent.
On this instance (pictured below), the RSS memory usage of dr_agent was over 6GB while fresh instances that had only been running for a few days had dr_agent using 250MB. The usage in memory grows by about 50MB per day (50MB x 30.5 days x 5 months = 7625 MB, which checks out). This is the case on other instances.
Any suggestions as to how to alleviate this?