This container contains a Dataloop (Outlyer) agent. It will create virtual agents in Outlyer for each running container. Depending on which OS you are running on your Docker hosts you may need to add different run options.
The list of metrics returned for each running containers can be found here.
DATALOOP_AGENT_KEY=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
DATALOOP_NAME=docker_container_name
docker run -d \
-e "DATALOOP_AGENT_KEY=${DATALOOP_AGENT_KEY}" \
-e "DATALOOP_NAME=${DATALOOP_NAME}" \
-v /var/run/docker.sock:/var/run/docker.sock:ro \
-v /proc:/rootfs/proc:ro \
-v /sys/fs/cgroup:/rootfs/sys/fs/cgroup:ro \
outlyer/agent:latest
If using an Amazon Linux AMI you will need to change the /cgroup volume mount location
DATALOOP_AGENT_KEY=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
DATALOOP_NAME=docker_container_name
docker run -d \
-e "DATALOOP_AGENT_KEY=${DATALOOP_AGENT_KEY}" \
-e "DATALOOP_NAME=${DATALOOP_NAME}" \
-v /var/run/docker.sock:/var/run/docker.sock:ro \
-v /proc:/rootfs/proc:ro \
-v /cgroup:/rootfs/sys/fs/cgroup:ro \
outlyer/agent:latest
CentOS 6 supports Docker to version 1.7, whereat the platform appears to have been deprecated.
You can still run this container, if you install cgroups and adjust the volume mapping for the cgroups directory.
sudo yum install -y libcgroup
CentOS 6 has cgroups under /cgroups. Run the container accordingly
DATALOOP_AGENT_KEY=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
DATALOOP_NAME=docker_container_name
docker run -d \
-e "DATALOOP_AGENT_KEY=${DATALOOP_AGENT_KEY}" \
-e "DATALOOP_NAME=${DATALOOP_NAME}" \
-v /var/run/docker.sock:/var/run/docker.sock:ro \
-v /proc:/rootfs/proc:ro \
-v /cgroup:/rootfs/sys/fs/cgroup:ro \
outlyer/agent:latest
This container has been tested to run on Docker >= 1.7
Please contact us on https://slack.outlyer.com or support[at]outlyer.com for help.
If you dont see any memory metrics in your containers you will need to enable memory accounting in cgroups. To do that just add some kernel command-line parameters: cgroup_enable=memory swapaccount=1. More info from the docker documentation.
When running the agent with the above docker run options, some features will be missing. Please run the container with --net host if you would like the following features:
- CPU load average for your containers
- Container network connections
- Host network metrics sent under your agent container
If you'd like to have host network metrics sent under your agent container (rather than your actual container's network metrics) and you cannot run your container with --net host, please follow these steps:
1. Mount your host's sys folder on your agent container (e.g. pass `-v /sys:/rootfs/sys` to `docker run`);
2. On your account's main overview page <https://app.outlyer.com/YOUR_ORG/YOUR_ACCOUNT/#/setup/overview>, click on "Setup Monitoring";
3. Click on "Plugins List", then on "base.py".
4. On the code editor, there should be a function called `check_netio`.
Replace it with the one here: <https://gist.github.com/alexdias/d51ef12d02d9e8f56721c48d52517f5c>
5. Click `Save changes` to have your base plugin updated.
If you are behind a proxy server, you can pass the standard proxy environment variables to the docker container to have data passed through:
docker run -e HTTP_PROXY=http://proxy:port....
This docker image will accept various other environment variables to allow you to customize the configuration for the dataloop-agent
DATALOOP_AGENT_KEY Required The API Key for your Dataloop account. Default: None
DATALOOP_NAME Optional A name for your agent to appear as in Dataloop. Default: dataloop
DATALOOP_TAGS Optional A comma separated list of tags to apply to the agent. Default: docker
DATALOOP_FINGERPRINT Optional You can pass an existing fingerprint if you want to keep your data association in Dataloop. Default None
DATALOOP_DEBUG Optional You can pass in the debug flag yes/no or true/false to add extra logging to the agent. Default no
None
If you want to modify the container then feel free to submit a pull request.
This container is also provided as a base container running a Dataloop agent. You can extend it with embedded plugins, or with Prometheus exporters, for example.
Based upon Alpine Linux and using Tini for process managment.