Frontend blocking requests when not in GCP

Describe the bug

a request against the frontend app when deployed in a non-GCP environment lead to a 2 minutes delay

To Reproduce

• deploy the whole project in a K8s cluster WITHOUT an API answering on 169.254.169.254:80. This can be achieved using a K8s cluster out of AWS or GCP, or, like me, using Cilium Network Policies to block such traffic
• curl against the frontend service… it will block for ~ 2 minutes

Logs

{"message":"Tracing disabled.","severity":"info","timestamp":"2022-01-28T18:43:52.381216858Z"}
{"message":"Profiling disabled.","severity":"info","timestamp":"2022-01-28T18:43:52.381283518Z"}
{"message":"starting server on :8080","severity":"info","timestamp":"2022-01-28T18:43:52.381944193Z"}
{"http.req.id":"ec621cfe-d9b8-4965-aa94-bfd8033c15c5","http.req.method":"GET","http.req.path":"/","message":"request started","session":"d5cf17c3-c869-4898-bb58-efa12d1fb2dd","severity":"debug","timestamp":"2022-01-28T18:44:18.181114709Z"}
{"currency":"USD","http.req.id":"ec621cfe-d9b8-4965-aa94-bfd8033c15c5","http.req.method":"GET","http.req.path":"/","message":"home","session":"d5cf17c3-c869-4898-bb58-efa12d1fb2dd","severity":"info","timestamp":"2022-01-28T18:44:18.181157082Z"}
{"http.req.id":"ec621cfe-d9b8-4965-aa94-bfd8033c15c5","http.req.method":"GET","http.req.path":"/","message":"ENV_PLATFORM is: aws","session":"d5cf17c3-c869-4898-bb58-efa12d1fb2dd","severity":"debug","timestamp":"2022-01-28T18:44:18.346517667Z"}
{"http.req.id":"ec621cfe-d9b8-4965-aa94-bfd8033c15c5","http.req.method":"GET","http.req.path":"/","message":"Failed to fetch the name of the cluster in which the pod is runningcompute: Received 503 `upstream connect error or disconnect/reset before headers. reset reason: connection failure`","session":"d5cf17c3-c869-4898-bb58-efa12d1fb2dd","severity":"error","timestamp":"2022-01-28T18:45:20.438611873Z"}
{"http.req.id":"ec621cfe-d9b8-4965-aa94-bfd8033c15c5","http.req.method":"GET","http.req.path":"/","message":"Failed to fetch the Zone of the node where the pod is scheduledcompute: Received 503 `upstream connect error or disconnect/reset before headers. reset reason: connection failure`","session":"d5cf17c3-c869-4898-bb58-efa12d1fb2dd","severity":"error","timestamp":"2022-01-28T18:46:21.832806289Z"}

Screenshots

Environment

AWS EKS 1.21

Rant

I know it’s a Google project, but could we ensure NO google specific code is executer if the platform is not GCP ?

[prune998]

thanks @NimJay for putting the PR up ! Can’t wait !

[NimJay]

FYI - The current workaround is to use version 0.3.1. We are working on a solution in the meantime.