From Asher's email:
"Looking over parsoid logs and the code that creates it, I think that
refactoring could help deliver some of what I'd like without adding any
complexity. Parsoid already logs what's being parsed, request urls sent to
varnish, and parse times in ms, written to a mix of stdout and stderr. The
formatting could be standardized and/or the data combined into an access
log which we could parse for stats. It also wouldn't take much to use
node's dgram module to either udplog this to the mediawiki log server, send
stats directly to a statsd daemon, or both. This could indeed be done by
an engineer outside of parsoid team."
Gabriel's response:
"If you spec a format that is easily consumable by tools you would like
to use, then I would be happy to change the few lines that do our
current minimalistic logging to fit it. Apart from regular request
logging and cache requests (currently logged for debugging purposes)
we'd like to keep the ability to log exceptions. It probably makes sense
to split things up into request and error logs at that point."
Version: unspecified
Severity: normal
Whiteboard: incident-report