When running on Docker versions 18.9.2+ (with the fix for CVE-2019-5736), the runC binary is copied into each container. This adds an additional ~12 MiB of memory use to all containers.

The MongoDB Exporter containers are created with a limit of 16Mi, which is not enough to fit both runC and the exporter. This then results in errors like this:

Error: failed to start container "prometheus-exporter": Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:424: container init caused "process_linux.go:390: setting cgroup config for procHooks process caused \"failed to write 16777216 to memory.limit_in_bytes: write /sys/fs/cgroup/memory/kubepods/pod0d665952-39ad-11e9-bae8-0a580a0f1d38/prometheus-exporter/memory.limit_in_bytes: device or resource busy\""": unknown

The MongoDB Operator would need to be patched to increase the memory limits, like the prometheus-operator was patched.