rmccomb-bnet / kraken-logging-fluentd-daemonset Goto Github PK

Dependencies for the Fluentd DaemonSet for Kubernetes logging. The docker image for this repo is located at: quay.io/samsung_cnct/fluentd_daemonset. Currently this component reads Docker logs from var/log/containers, filters Kubernetes metadata and writes to a zookeeper/Kafka component.

kubectl create -f fluentd-daemonset.yaml

This filter adds the following data into the body of the log.

• namespace
• pod id
• pod name
• labels
• host
• container name
• docker container id

For more information on the filter or to see a list of configuration options: https://github.com/fabric8io/fluent-plugin-kubernetes_metadata_filter

This filter will add the filename and filepath to the event metadata. Requires setup of shared-logging-directory: https://github.com/samsung-cnct/shared-logging-directory

There is an optional script in the init directory to monitor resource usage for Fluentd running in your cluster. By modifying your Dockerfile the script will pull CPU and memory consumption constantly. To use the script, replace the last line in the Dockerfile with:

CMD ["/start.sh"]

Rebuild your docker image and redeploy the daemonset with new docker image. The script will run, you can stress-test your logging system and when you want to examine the data you can pipe the logs from your daemonsets into a file (on your local machine) with:

$ kubectl logs fluentd-daemon-example >> logs.dat

Grep the file for your metrics with:

$ grep "\[metrics\]" <logs.dat> | cut -d" " -f2,3 > <clean.dat>

You will have two columns, CPU & MEM, which you can graph or pinpoint peak usage related to your stress testing.

GNUPLOT is an easy way to quickly visualize the data. Start GNUPLOT (install if necessary) and run:

gnuplot> plot '<clean.dat>' using 1 with lines for CPU usage gnuplot> plot '<clean.dat>' using 2 with lines for MEM usage