Guide for running on K8s

[alexdebrie]

Event Gateway is attractive to Kubernetes users (example: Kelsey Hightower's KubeCon demo).

This ticket is to add documentation and supporting scripts to make it easy to get started with Event Gateway on K8s.

[fargozhu]

The cpu issue was solved by removing resources definition but now I am facing with pod crashes with the following error message:

2018-06-27 09:37:07.726410 I | etcdserver/api/v3rpc: grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp 127.0.0.1:2379: getsockopt: connection refused"; Reconnecting to {127.0.0.1:2379 }
2018-06-27 09:37:08.726630 I | etcdserver/api/v3rpc: grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp 127.0.0.1:2379: getsockopt: connection refused"; Reconnecting to {127.0.0.1:2379 }
2018-06-27 09:37:10.055508 I | etcdserver/api/v3rpc: grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp 127.0.0.1:2379: getsockopt: connection refused"; Reconnecting to {127.0.0.1:2379 }
2018-06-27 09:37:12.524521 I | etcdserver/api/v3rpc: grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp 127.0.0.1:2379: getsockopt: connection refused"; Reconnecting to {127.0.0.1:2379 }
2018-06-27 09:37:17.216986 I | etcdserver/api/v3rpc: grpc: addrConn.resetTransport failed to create client transport: connection error: desc = "transport: dial tcp 127.0.0.1:2379: getsockopt: connection refused"; Reconnecting to {127.0.0.1:2379 }
2018-06-27 09:37:17.726471 I | etcdserver/api/v3rpc: Failed to dial 127.0.0.1:2379: context canceled; please retry.
{"level":"fatal","ts":1530092237.7264407,"msg":"Cannot create KV client.","error":"dial tcp 127.0.0.1:2379: getsockopt: connection refused"}

[alexdebrie]

Hey @fargozhu, do you have an etcd cluster running?

[fargozhu]

Hi @alexdebrie , I do not...

[baniol]

Hi,
all the examples from the Kelsey Hightower's KubeCon demo repo work all right for me.
For an easy start, (apart from the k8s yaml files) a Helm chart, with etcd-operator as a dependency, would be a good thing (I could handle this).
However, stepping through the examples, I've come across a security issue. The service script included in the demo spins up a load balancer with a public endpoint. Now, if you run serverless deploy command with this endpoint in the config, it registers functions with your local AWS credentials, which then are publicly exposed via /v1/spaces/default/functions.