Rather than needing to dig these out of[[ https://wikitech.wikimedia.org/wiki/Analytics/Data_Lake/Traffic/Webrequest | Hadoop ]] it would be easier if we logged these to logstash.
The service-runner provides a configured[[ https://github.com/wikimedia/service-runner | logging object ]] which will write to [[ https://wikitech.wikimedia.org/wiki/Logstash | logstash ]]. We should use this to log the failed requests for:
!! For discussion: Should we also log abnormally long-running requests!!