2020-10-28 01:20:01 UTC - Ning yougang: @Jiang PengCheng,if changed to the nginx configuration like you said, if the controller pods are restarted, need to restart the nginx pods as well, otherwise the nginx can't recognize upstream servers. reported below error. e.g. ```2020/10/27 10:46:03 [error] 42#42: *3130948 connect() failed (113: Host is unreachable) while connecting to upstream, client: <http://xxx.xxx.xxx.xxx|xxx.xxx.xxx.xxx>, server: ~^(?<namespace>[0-9a-zA-Z-]+)\.$, request: "POST /api/v1/namespaces/whisk.system/actions/hello1?blocking=true&result=true HTTP/1.1", upstream: "<http://10.171.184.47:8080/api/v1/namespaces/whisk.system/actions/hello1?blocking=true&result=true>", host: "<http://xxx.xxx.xxx.xxx|xxx.xxx.xxx.xxx>"``` https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1603848001170100 ---- 2020-10-28 01:25:11 UTC - Ning yougang: hm..BTW, if i want to use above configuration, e.g. ``` upstream controllers { # Mark the controller as unavailable after fail_timeout seconds, to not get any requests during restart. # Otherwise, nginx would dispatch requests when the container is up, but the backend in the container not. # From the docs: # "normally, requests with a non-idempotent method (POST, LOCK, PATCH) are not passed to # the next server if a request has been sent to an upstream server" server <http://openwhisk-controller-0.jiangpengcheng.svc.example.com;:8080|openwhisk-controller-0.jiangpengcheng.svc.example.com;:8080> fail_timeout=60s; server <http://openwhisk-controller-1.jiangpengcheng.svc.example.com;:8080|openwhisk-controller-1.jiangpengcheng.svc.example.com;:8080> fail_timeout=60s; server <http://openwhisk-controller-2.jiangpengcheng.svc.example.com;:8080|openwhisk-controller-2.jiangpengcheng.svc.example.com;:8080> fail_timeout=60s; keepalive 512; }``` And when controller pods are restarted, how to make nginx pods can access its own upstream backend servers without restart nginx pods? https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1603848311171800 ---- 2020-10-28 01:42:40 UTC - Ning yougang: I did another experiment, after restarted controller pods, i ping the pod dnsName in nginx pod(e.g. ping <http://openwhisk-controller-0.jiangpengcheng.svc.example.com|openwhisk-controller-0.jiangpengcheng.svc.example.com>), access successfully https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1603849360173600 ---- 2020-10-28 06:31:10 UTC - Michele Sciabarra: thanks that will be much apprecviated https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1603866670173800?thread_ts=1603551396.137600&cid=C3TPCAQG1 ----
