> With nginx->FastRouter->nodes and current uWSGI master branch I see: > > ab - 10000 requests done, including 803 failed > node - 9197 requests done, 0 write_errors > fastrouter - 9950 requests, 50 failcnt > > logs on fastrouter shows: > > 747 x fr_instance_read(): Connection reset by peer > [plugins/fastrouter/fastrouter.c line 127] > > 50 x [uwsgi-fastrouter] unable to connect() to node "172.16.200.56:3001": > Connection timed out > looks like above is counted in "failcnt"
hmm, 6 errors lost here... Maybe some error in the node side ? > > Also I've noticed that "ref" shows a lot more than my node (or its > backlog) > can handle, high ref value stays there for a while, probably until request > gets a timeout or read error. yes, you are experiencing connection timeouts (maybe for the full listen queue) so the ref will be accounted until that timeout. The important thing is that ref come back to 0 after that timeout. -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
