We manage logs of most PMM Server components with supervisord. Currently, there are three exceptions: nginx, PostgreSQL, and ClickHouse, who manage their logs themselves with configurations in three different places.
We should change their configuration to log to stderr, and manage their logs with supervisord as for every other component.
Should we do it via Ansible templates in pmm-update instead of sed? (see https://github.com/Percona-Lab/pmm-submodules/blob/f602d81bba2e94e2664ad63a2206ed9052df899a/build/bin/build-server-rpm#L188)