Error running a development Kubernetes cluster with gulp task

[satoshi75nakamoto]

Issue details

$ gulp local-up-cluster
[11:08:18] Requiring external module babel-register
[11:08:20] Using gulpfile ~/code/dashboard/gulpfile.babel.js
[11:08:20] Starting 'spawn-cluster'...
[11:08:20] Starting 'wait-for-cluster'...
[11:08:20] Starting 'wait-for-heapster'...
[11:08:21] Finished 'spawn-cluster' after 501 ms
[11:08:21] Waiting for a Kubernetes cluster at http://localhost:8080...
[11:08:21] Waiting for a Heapster ...
undefined:1
404: Page Not Found
   ^

SyntaxError: Unexpected token :
    at Object.parse (native)
    at cluster.js:73:27
    at ChildProcess.exithandler (child_process.js:193:7)
    at emitTwo (events.js:100:13)
    at ChildProcess.emit (events.js:185:7)
    at maybeClose (internal/child_process.js:850:16)
    at Process.ChildProcess._handle.onexit (internal/child_process.js:215:5)

Environment

Dashboard version: commit 6937a5a5f3f810f862b41bba8edde3e0eade2200
Kubernetes version: 1.1.2
Operating system: Ubuntu 15.10
Node.js version: v5.9.0
Go version: go version go1.6 linux/amd64

Steps to reproduce

Observed result

Expected result

Comments

[satoshi75nakamoto]

I can confirm that heapster:v0.20.0-alpha9 is running.

curl http://localhost:8082/api/v1/model/debug/allkeys
[
  "node:127.0.0.1/container:docker-daemon",
  "node:127.0.0.1/container:docker/f554eca21bfea905edb4ea221423921a9ff7943449bf127154afbc3fbef60160",
  "node:127.0.0.1/container:user.slice/user-1000.slice",
  "node:127.0.0.1/container:user.slice/user-1000.slice/session-8.scope",
  "node:127.0.0.1/container:user.slice/user-1000.slice/[email protected]",
  "node:127.0.0.1/container:docker/f554eca21bfea905edb4ea221423921a9ff7943449bf127154afbc3fbef60160/kube-proxy",
  "node:127.0.0.1/container:docker/f16be901c68ba23f11154255a1e7a608ea5b31d7fbda0e34e159149ca938d86b",
  "node:127.0.0.1/container:docker/854011cdfbb24a70d8b8250b19b5676560fc3321dd86be3c65353e7ab9c1ed65",
  "node:127.0.0.1/container:user.slice",
  "node:127.0.0.1/container:docker/f7d97d4b8a701d4b23671c45cf85b12d231f8ae89f7c2f476e6147386900d205",
  "node:127.0.0.1/container:user.slice/user-1000.slice/session-5.scope",
  "node:127.0.0.1/container:user.slice/user-1000.slice/session-2.scope",
  "node:127.0.0.1/container:docker/5bf9679f6265a33c98b69af79a0bb220f635cef18eae3d0f3c56a24e10f61273",
  "node:127.0.0.1/container:docker/fbef807806da182add1aac060dfbf83de7d006025b97c15a340f6686a192ba5a",
  "node:127.0.0.1/container:docker/6b9c5873e922bf1d05b13dab1b7a4626d5c6655cae9a5b262661d60e87186156",
  "node:127.0.0.1/container:system.slice",
  "cluster",
  "node:127.0.0.1/container:docker",
  "node:127.0.0.1/container:docker/93aef6a2ebbb39662aeda2f70a4d8f49ce5a13315a14d33653a65366fc310d11",
  "node:127.0.0.1/container:user.slice/user-1000.slice/session-3.scope",
  "node:127.0.0.1",
  "node:127.0.0.1/container:kubelet"

[satoshi75nakamoto]

I don't seem to find that endpoint in: https://github.com/kubernetes/heapster/blob/master/docs/storage-schema.md

[cheld]

local-up-cluster is launching Kubernetes and Heapster and waiting for both to become alive. The problem could also be related to API server as gulp executes both tasks in parallel.

Hmm...I have an idea....are you running something already on port 8080? Gulp is waiting for Kubernetes to respond on port 8080.

BTW: you can run the containers directly with the script build/hypercube.sh - might simplify your debugging attempts.

[satoshi75nakamoto]

@cheld thanks for the assistance but it appears that nothing else is running on port 8080 and the services successfully start as expected. The issue is that the script is unable to parse the 404: Page Not Found error that is encountered from heapster:v0.20.0-alpha9 which is running on http://localhost:8082 it appears that the script is attempting to access an endpoint that doesn't exist. Do you mind trying a fresh checkout at commit 6937a5a5f3f810f862b41bba8edde3e0eade2200 and attempt to run $ gulp local-up-cluster.

I'm aware that I can run build/hypercube.sh directly. This issue is more about making sure that the project works as described in the https://github.com/kubernetes/dashboard/blob/master/README.md

[cheld]

Ah ok, I see. But I cannot reproduce - kind of. The cluster starts without an error message. But a manual check of the health-check url fails...I will investigate...

[bryk]

@cheld I'm assigning this to you. Please work with @preillyme, because we really want to have dev instructions that actually work :)

[cheld]

I can reproduce now with the following steps:

gulp local-up-cluster
docker kill $(docker ps -q)
gulp local-up-cluster

[cheld]

The problem was caused by:

1. Incorrect error handling
2. The health check URL was not updated after upgrading Heapster

Fixed