Bug Report

• Concourse version: 3.11.0
• Deployment type (BOSH/Docker/binary): Docker
• Infrastructure/IaaS: ubuntu VM
• Browser (if applicable): Chrome?
• Did this used to work? Yes

Beginning with concourse v3.11.0 we observe an almost linear rate of memory growth in the concourse-web container.

Below is a grafana graph showing the host-level metrics over ~24h period:

Below is a netdata graph of the host showing the concourse-web container and associated memory growth over ~1hr period:

Got access to the host in question, and this is what glances shows to confirm the narrative of the web growth story:

It may be related to the fact that we have dashboards running on monitors showing the pipeline continually. When looking at a different installation not running a continuous dashboard/monitor, the same type of memory growth is not observed.

Raising this issue because two other people in discord mentioned seeing the same thing or something similar.