docker-ee-chargeback

This repo contains a utility tool to generate hourly reports of usage against a Docker EE deployment, which can then be used to bill internal departments for usage.

This tool exposes the Collection for the given resources, which can then be mapped back to internal teams or billing departments. We recommend different billable departments have their resources placed in the different collections, or consider using hierarchies to separate to help in structuring a cleaner chargeback structure.

The key metrics exposed in this tool are:

• CPU - Cumulative CPU second used over the duration of the report (typically 1 hour)
• Memory - Min, max, and average MB of memory used by the container during the interval
• Network - RX/TX MB for each container
• Container Storage - Min, Max, and average MB of ephemeral usage within the container (excluding the image layers)
• Volume - Min, Max, and average MB used by storage volumes on the system

Here are a few examples:

TYPE,COLLECTION,CONTAINER ID,NAME,SAMPLE DURATION IN SECONDS,CUMULATIVE VALUE,MIN VALUE,MAX VALUE,AVE VALUE
CPU SECONDS,/dev/sneaky,bc1abd880e77,sneaky1.1.3b09xmyaoy4niw29xd0h004ig,480.000000,273.107326,0.000000,0.000000,0.000000
CPU SECONDS,/prod,b146e7f333c5,prodweb1.1.ec0hf2beeq2kvcm7mspeb8ucg,3600.000000,0.000000,0.000000,0.000000,0.000000
MEM MB,/dev/sneaky,bc1abd880e77,sneaky1.1.3b09xmyaoy4niw29xd0h004ig,480.000000,0.000000,145.011719,145.023438,145.019965
MEM MB,/prod,b146e7f333c5,prodweb1.1.ec0hf2beeq2kvcm7mspeb8ucg,3600.000000,0.000000,1.410156,1.410156,1.410156
NETWORK RX MB,/dev/sneaky,bc1abd880e77,sneaky1.1.3b09xmyaoy4niw29xd0h004ig,480.000000,0.001236,0.000000,0.000000,0.000000
NETWORK RX MB,/prod,b146e7f333c5,prodweb1.1.ec0hf2beeq2kvcm7mspeb8ucg,3600.000000,0.000000,0.000000,0.000000,0.000000
NETWORK TX MB,/dev/sneaky,bc1abd880e77,sneaky1.1.3b09xmyaoy4niw29xd0h004ig,480.000000,0.000349,0.000000,0.000000,0.000000
NETWORK TX MB,/prod,b146e7f333c5,prodweb1.1.ec0hf2beeq2kvcm7mspeb8ucg,3600.000000,0.000000,0.000000,0.000000,0.000000
VOLUME MB,/stage,,192.168.122.105:12376/bunchostuff,300.000000,0.000000,529.484172,529.484172,529.484172
CONTAINER STORAGE MB,/stage,e01c601c518f,builder.1.mm39xsyro9scyoo5rbb11nypk,300.000000,0.000000,278.255743,278.255743,278.255743

Using published images from Docker Hub

Note: The metrics stored within EE vary across releases. You should use a version tag of this tool that matches the major.minor version of your UCP.

1. Download an admin bundle from UCP on your EE cluster
2. Source the env script so your docker commands are sent through UCP

Run manually to verify you're able to get the CSV results.

MANAGER=$(docker node ls --filter role=manager --format '{{.Hostname}}' | head -1)
docker run --rm -e constraint:node==${MANAGER} -v ucp-node-certs:/certs \
    dhiltgen/chargeback:2.2 --certs /certs \
    --ucp $(echo $DOCKER_HOST | cut -f3 -d/ | cut -f1 -d:)

Once you've confirmed this looks good, you can set up a cron-job to run this command hourly against your cluster. You can append the results to a single file and omit the header row with --omit-header added at the end of the command. Run docker run --rm dhiltgen/chargeback:2.2 --help for more detailed usage information.

Developing

For Developers wishing to modify this tool

If you're building against a UCP cluster, since the image will ultimately need to run on a manager, you can use scheduler constraints to build on a manager (not recommended for a production cluster!)

Once you've downloaded a bundle (preferably admin, since you'll need an admin bundle to run the final image anyway) you can build it with:

MANAGER=$(docker node ls --filter role=manager --format '{{.Hostname}}' | head -1)

docker build --build-arg constraint:node==${MANAGER} -t chargeback:2.2 .

Then to run it:

docker run --rm -e constraint:node==${MANAGER} -v ucp-node-certs:/certs \
    chargeback:2.2 --certs /certs \
    --ucp $(echo $DOCKER_HOST | cut -f3 -d/ | cut -f1 -d:)