InfluxDB appears to not be tracking all data
It appears InfluDB is not tracking all data, or somehow not all data is sent to it. For example, the dashboard for API calls to /api/v3/internal/allowed
(http://performance.gitlab.net/dashboard/db/grape-allowed-ssh-requests) only displays several requests per second whereas Unicorn log files show a far greater amount of requests.
So far changing net.core.rmem_max
and net.core.netdev_max_backlog
(on the workers and the InfluxDB server) has had no effect.
cc @pcarranza