Hello!

The following patch introduces less intrusive upstream server recheck after
server was considered dead.

Specifically, only one request is allowed once fail_timeout passes, and
upstream server is only marked "live" once this request completes
successfully.

I think it improves things, though there are some concerns related to
"long streaming" workloads where request may not complete in a reasonable
time (and hence backend won't be reintroduced into upstream cluster
for a relatively long time). I tend to think this is acceptable, but I
would like to hear some comments from people with such workloads.

Maxim Dounin

_______________________________________________
nginx-devel mailing list
nginx-devel@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx-devel