DEA reports running but cc says down.

[Harry Zhang]

I hava a app deployed on CFv2 but I always got 0% state of this app.

So I checked the log of HM and CC, I found the the DEA has sent heartbeat like this which means the app is running:

heartbeat: {:droplets=>[{:cc_partition=>"ng", :droplet=>"2224cde2-56b7-40a2-aadf-5fb02320ee5b", :version=>"ae3c8a1c-9d24-4607-a3e9-c300d8e9ded8", :instance=>"21d01f29e82b02c7f69affa72ed17cdf", :index=>0, :state=>"RUNNING", :state_timestamp=>1377676624.474158}, {:cc_partition=>"ng", :droplet=>"2224cde2-56b7-40a2-aadf-5fb02320ee5b", :version=>"ae3c8a1c-9d24-4607-a3e9-c300d8e9ded8", :instance=>"88f9e85ff464c9bce98611a25439a2e8", :index=>1, :state=>"RUNNING", :state_timestamp=>1377676624.3643758}], :dea=>"0-6d51dc59507852e87668ae9070886b0a", :prod=>false}

But, "cf health myapp" returns another state:

Getting health status>>>
REQUEST: GET http://api2.xxx.xxx/v2/apps/2224cde2-56b7-40a2-aadf-5fb02320ee5b/instances
REQUEST_HEADERS:
  Accept : application/json
  Authorization : [PRIVATE DATA HIDDEN]
  Content-Length : 0.  
   RESPONSE: [200]
RESPONSE_HEADERS:
  content-length : 81
  content-type : application/json;charset=utf-8
  date : Wed, 28 Aug 2013 09:40:02 GMT
  server : thin 1.5.1 codename Straight Razor
  x-content-type-options : nosniff
  x-vcap-request-id : 34261f06-e600-4a60-80c3-5c6083ddb6ca
RESPONSE_BODY:
{
  "0": {
    "state": "DOWN",
    "since": 1377682802
  },
  "1": {
    "state": "DOWN",
    "since": 1377682802
  }
}
<<<
... OK

I'm really confused now about these states. Could any give me tips where should I begin to check, believe CC, or believe DEA? Why CC consider my app as down?

[Cornelia Davis]

Please check the values for cc_partition in your cloud_controller_ng.yml (/var/vcap/jobs/cloud_controller_ng/config) and health_manager_next.yml (/var/vcap/jobs/health_manager_next). The two values need to match, though I'm betting you'll find that one is set to "default" and the other is set to "ng". I've filed a pull request to have these values match by default but in the mean time you can set them to match by setting one of them in your deployment manifest as follows:

  cc: &cc

    logging_level: debug

    external_host: ccng

    cc_partition: default

    ...

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

[Harry Zhang]

Thank you Cornelia, I checked that part, they are all "ng" ... Any more tips for this case? Or could u plz help me with how CC and HM get the state value?

[Harry Zhang]

Oh, resolved! The 0.4.26 nats client can not work well with 0.5.0-beta nats server. I'll keep looking it and contact with Derek.

On Wednesday, August 28, 2013 9:09:02 PM UTC+8, Cornelia Davis wrote:

[Cornelia Davis]

Interesting.  Thanks for reporting back your findings Harry!