I'm trying to setup logging on my GKE cluster by following the suggestions 
from this thread 
<https://groups.google.com/forum/#!searchin/kubernetes-users/fluentd|sort:relevance/kubernetes-users/Q1nvl8IAbqc/BXOZNBUWAgAJ>,
 
but have hit a wall.

Specifically, I've:

1) Created a new cluster on GKE with google logging disabled
2) Ran `kubectl create -f ...` on each of the following 5 components:

# From: 
https://github.com/kubernetes/kubernetes/tree/dae5ac482861382e18b1e7b2943b1b7f333c6a2a/cluster/addons/fluentd-elasticsearch
es-controller.yaml
es-service.yaml
kibana-controller.yaml
kibana-service.yaml

# From: 
https://github.com/kubernetes/kubernetes/blob/714db746118923a3918e28aacb4564e91afdd368/cluster/saltbase/salt/fluentd-es/fluentd-es.yaml
fluentd-es.yaml

However I'm seeing some odd results. My service/rc/pods for kibana and 
elastic search are created, then die off very quickly. 

Here's what it looks like before those components die off:

$kubectl get all --namespace=kube-system
NAME                                                           DESIRED     
   CURRENT       AGE
elasticsearch-logging-v1                                       2           
   2             4s
kibana-logging-v1                                              1           
   1             3s
kube-dns-v17.1                                                 2           
   2             2h
kubernetes-dashboard-v1.1.1                                    1           
   1             2h
l7-default-backend-v1.0                                        1           
   1             2h
NAME                                                           CLUSTER-IP   
  EXTERNAL-IP   PORT(S)         AGE
default-http-backend                                           10.3.245.50 
   <nodes>       80/TCP          2h
elasticsearch-logging                                          10.3.242.22 
   <none>        9200/TCP        5s
heapster                                                       10.3.246.252 
  <none>        80/TCP          2h
kibana-logging                                                 10.3.241.94 
   <nodes>       5601/TCP        3s
kube-dns                                                       10.3.240.10 
   <none>        53/UDP,53/TCP   2h
kubernetes-dashboard                                           10.3.242.135 
  <none>        80/TCP          2h
NAME                                                           READY       
   STATUS        RESTARTS        AGE
elasticsearch-logging-v1-lq5gx                                 1/1         
   Running       0               4s
elasticsearch-logging-v1-p78mn                                 1/1         
   Running       0               4s
fluentd-elasticsearch                                          1/1         
   Running       0               6s
heapster-v1.1.0-2096339923-4j1vu                               2/2         
   Running       0               2h
kibana-logging-v1-km83q                                        1/1         
   Running       0               3s
kube-dns-v17.1-aqteq                                           3/3         
   Running       0               2h
kube-dns-v17.1-qcm9z                                           3/3         
   Running       0               2h
kube-proxy-gke-pachyderm-log-test-default-pool-8429ab58-1sio   1/1         
   Running       0               2h
kube-proxy-gke-pachyderm-log-test-default-pool-8429ab58-b7tw   1/1         
   Running       0               2h
kube-proxy-gke-pachyderm-log-test-default-pool-8429ab58-eead   1/1         
   Running       0               2h
kubernetes-dashboard-v1.1.1-aajls                              1/1         
   Running       0               2h
l7-default-backend-v1.0-filol                                  1/1         
   Running       0               2h

Normally I'd get the logs using the previous flag, but that doesn't seem to 
work. However if I get the logs fast enough, I can see a few things, but 
nothing out of the ordinary.

$kubectl logs elasticsearch-logging-v1-lq5gx --namespace=kube-system
I0823 21:16:06.814369       5 elasticsearch_logging_discovery.go:42] 
Kubernetes Elasticsearch logging discovery
I0823 21:16:07.829359       5 elasticsearch_logging_discovery.go:75] Found [
"10.0.1.4" "10.0.2.6"]

Which looks pretty normal to me. Grabbing the logs from the kibana pod 
quickly, I see:

$kubectl logs kibana-logging-v1-km83q --namespace=kube-system
ELASTICSEARCH_URL=http://elasticsearch-logging:9200
{"@timestamp":"2016-08-23T21:16:10.223Z","level":"error","node_env":
"production","error":"Request error, retrying -- connect ECONNREFUSED"}
{"@timestamp":"2016-08-23T21:16:10.227Z","level":"warn","message":"Unable 
to revive connection: http://elasticsearch-logging:9200/","node_env":
"production"}
{"@timestamp":"2016-08-23T21:16:10.227Z","level":"warn","message":"No 
living connections","node_env":"production"}
{"@timestamp":"2016-08-23T21:16:10.229Z","level":"info","message":"Unable 
to connect to elasticsearch at http://elasticsearch-logging:9200. Retrying 
in 2.5 seconds.","node_env":"production"}

Which makes sense ... considering the elastic search pod dies off very 
quickly.

The fluentd pod reports an error connecting to elastic search, but again, 
thats not surprising considering it quickly dies off.

Clearly, I'm missing some steps in configuring these services to connect to 
each other. Any advice along these lines is appreciated.

I know that I should probably be using a daemon set to spin up the fluentd 
pod, but I thought I could get these components wired together before 
worrying about that step. Maybe thats not the case. Either way, I'd 
appreciate any pointers in setting that up as well.

-- 
You received this message because you are subscribed to the Google Groups 
"Kubernetes user discussion and Q&A" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/kubernetes-users.
For more options, visit https://groups.google.com/d/optout.
  • [kubernetes-... Sean Jezewski
    • Re: [ku... 'Vishnu Kannan' via Kubernetes user discussion and Q&A
      • Re:... Sean Jezewski
        • ... 'Piotr Szczesniak' via Kubernetes user discussion and Q&A
        • ... Sean Jezewski

Reply via email to