Google has an Open Source (Apache License 2.0) solution for instant monitoring docker containers, the Container Advisor (cAdvisor). It is a powerful simple solution, for instant monitoring. The drawback, it keeps no history. You may export the data to Elasticsearch, but keep in mind, it is an extra effort and not current post subject. I’m going to introduce you basic problems with docker in this post.
I did a little Docker container - Elasticsearch Curator (
es-curator) - and wanted to know how much memory does it consume on the big machine. Therefore you might use
docker stats. It might give you a very long list like this (output shortened):
CONTAINER CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O 0220fea60e6a 0.08% 720.9 MB / 134.3 GB 0.54% 54.06 MB / 45.96 MB 128.8 MB / 438.3 kB .. dbfc1c6bf14a 0.22% 900.4 MB / 134.3 GB 0.67% 166.3 MB / 70.31 MB 152.8 MB / 25.99 MB e21a72e127ed 0.95% 1.664 GB / 134.3 GB 1.24% 150.9 MB / 236.6 MB 158.3 MB / 39.27 MB e62122b6c9f6 22.74% 8.224 GB / 8.59 GB 95.74% 0 B / 0 B 88.09 GB / 183.2 GB
Holy shit, I don’t know my container id, so therefore I need to look it up with
docker ps | grep es-curator.
Then I asked myself, there must be someone out there, who did already the monitoring job. That’s cAdvisor. Just run the docker run command and it works like that.
The overview page, gives you nice gauges:
As you can see, Google Charts is behind the curtain, and a good indicator for problem areas.
If I do on my shell (output prettified), I can look up my process ids.
# docker top es-curator UID PID PPID C STIME TTY TIME CMD root 27446 17307 0 07:15 ? 00:00:00 /bin/sh /docker-entrypoint.sh root 27662 27446 0 07:15 ? 00:00:00 crond -f d8 -l 8
This is displayed the process overview:
Having 32 CPUs can be a little bit messy in the diagram .
You can filter with cAdvisor the respective Interface.
But it much simpler if you open the overview of the desired container .