> Hi all, > > Few months ago I have some 502 broken pipe errors in high load nginx+uwsgi > server. > > connect() to unix:/tmp/xxxx.sock failed (11: Resource temporarily > unavailable) while connecting to upstream > > But uWSGI does *NOT* report something like > > *** uWSGI listen queue of socket 8 full !!! *** > > Turns out Unix domain sockets does not count backlog connections. Only TCP > has TCP_INFO. And it has to use a kernel patch called SIOBKLGQ. (the > get_linux_unbit_SIOBKLGQ() in master.c) > > > http://lists.openwall.net/netdev/2012/03/16/11 > > Today I stumbled upon this: > > https://kernel.googlesource.com/pub/scm/linux/kernel/git/shemminger/iproute2/+/defd61ca9191e0c93950e7ea5466682a67dcbafb%5E%21/#F0 > > You can use netlink to detect domain socket queue length, does this help > with the situation? > _______________________________________________ > uWSGI mailing list > [email protected] > http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi >
Get latest code (2.0-rc1 or github) build it and the install this plugin: https://github.com/unbit/uwsgi-netlink uwsgi --build-plugin uwsgi-netlink then just add "--plugin netlink" to let it manage unix sockets -- Roberto De Ioris http://unbit.it _______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
