I think we need keep the logs for each benchmark that we run. We probably want something similar to what Travis CI provides. A live stream of logs and eventually store the logs some where.
The motivation stems from the fact that the logs are not being saved and as a result I did noticed that Discourse Ruby trunk benchmarks were failing due to a failed Gem install.
Travis implements GitHub - travis-ci/travis-worker: [DEPRECATED] This project is deprecated in favor of travis-ci/worker.