The server daemon logs to standard output and error. With multiple Parsoid server instances, you end up with logs on various machines and it might fill in the local disk.
The daemon should support logging via syslog, ideally relaying directly to a given syslog IP.
Version: unspecified
Severity: normal
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | • marcoil | T51762 Logging: emit structured logs via node-bunyan | |||
| Resolved | None | T59711 Logging: server should log to logstash | |||
| Resolved | ssastry | T55723 Package Parsoid and clean up deploy system |
Alternatively we can leave this to systemd or upstart, with the added bonus of automatic log rotation. See bug 53723 (packaging / systemd / upstart) and bug 49762 (tweak logging).
We now have bunyan support. Production should probably log to logstash via gelf, similar to OCG.
The original idea being my request was to get all logs centralized somewhere (i.e. syslog). Now that all Parsoid instances send to a central logstash in both production and beta, that fulfill the request.
I confirmed the Parsoid logs are properly sent on the beta cluster logstash ( http://logstash-beta.wmflabs.org/ )
Thank you everyone!