2010/9/10 Roberto De Ioris <[email protected]>: > > As having -L in production is a common thing, we > should think of a "list of rule" that mark requests for logging > (even if logging is disabled) > > For now i have thinked: > > --log-slow-requests <N> (requests that takes > N milliseconds will be logged) > --log-zero-headers (log requests that send zero headers) > --log-http-status <X,Y,Z,...> (log requests with specific http status > response) > --log-empty-body (log requests with an empty body) > > Others ?
Many facts can already be retrieved from the downstream webserver's logs. I would only log requests that are causing errors in uwsgi, such as harakiri and broken pipe. This is the output of uwsgi on my webserver, I can't figure out which requests broke. *** HARAKIRI ON WORKER 2 (pid: 31152) *** DAMN ! process 31152 died :( trying respawn ... Respawned uWSGI worker (new pid: 32636) *** HARAKIRI ON WORKER 2 (pid: 32636) *** DAMN ! process 32636 died :( trying respawn ... Respawned uWSGI worker (new pid: 32641) *** HARAKIRI ON WORKER 3 (pid: 31153) *** DAMN ! process 31153 died :( trying respawn ... Respawned uWSGI worker (new pid: 10638) *** HARAKIRI ON WORKER 2 (pid: 32641) *** DAMN ! process 32641 died :( trying respawn ... Respawned uWSGI worker (new pid: 10698) *** HARAKIRI ON WORKER 3 (pid: 10638) *** DAMN ! process 10638 died :( trying respawn ... Respawned uWSGI worker (new pid: 10728) *** HARAKIRI ON WORKER 3 (pid: 10728) *** DAMN ! process 10728 died :( trying respawn ... Respawned uWSGI worker (new pid: 12570) SIGPIPE: writing to a closed pipe/socket/fd !!! write(): Broken pipe [pyutils.c line 101] SIGPIPE: writing to a closed pipe/socket/fd !!! write(): Broken pipe [pyutils.c line 101] -- Jean-Baptiste Quenot _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
