Logging to AWS Elasticsearch Service from Kubernetes
In this guide we’re going to setup Amazon’s Elasticsearch service and forward logs from our Kubernetes cluster to it.
We will be using fluentd with the “aws_elasticsearch_plugin” to accomplish this.
Step 1: Create an IAM user called: “elasticsearch” (chose: AWS Programatic Access) and download the credentials.
Step 2: Create an Amazon Elasticsearch Service instance (Dashboard: Services->Analytics->Elasticsearch Service). For this example were only going to use a single instance. Feel free to choose whatever fits you best.
Note: if your HTTP payloads will be larger than 10mb than the smallest instance size you can use is: m3.xlarge
Step 3: Modify the Policy of the instance to allow both the “elasticsearch” user as well as all the NAT gateways of your Kubernetes cluster (this is assuming private-topology) and also any IP address that you wish to access the Kibana dashboard from.
Example policy below:
Step 4: Modify the fluentd-daemonset.yml below to add the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY for the elasticsearch user you made in Step 1. and a few other variables (AWS_REGION, AWS_ELASTICSEARCH_URL (This is going to be the endpoint thats generated once the instance is made).
Step 5: Launch the DaemonSet into kubernetes:
kubectl create -f fluentd-aws-es-daemonset.yml
Logs should now be flowing from all pods into Amazon Elasticsearch Service.
You can browser to your Kibana endpoint url and take a look: search-<name>-xxxxxx.us-east-1.es.amazonaws.com/_plugin/kibana/
Thanks to https://github.com/cheungpat for providing the docker image containing fluentd and the aws elasticsearch plugin: