Awesome
Upkick
Unattended upgrades for Docker containers, the hard way.
Installing
$ go get github.com/camptocamp/upkick
What does it do?
Upkick helps you keep your containers up-to-date. When you launch it, it performs the following actions:
- list all running containers on the Docker socket
- update (pull) all images used in the containers
- unless
--warn-only
is used, stop/remove all containers using outdated images (you need an orchestrator such as Rancher to restart them) - if a Prometheus gateway is provided, push metrics to it
Isn't that what catalog templates are for?
No. Catalog templates usually use tagged versions of images. However, Docker tags are not fixed: they correspond more to branches actually. So while a new catalog template might upgrade the tag and force an upgrade of a container, there is never a garantee that the container runs on the latest hash for the given tag. This is a concern for security, as images should be rebuilt on a regular basis.
Rancher allows to set containers to "Always pull image before creating". While this is useful, it only garantees that images are updated when containers are recreated. We want containers to be up-to-date all the time!
Usage
Usage:
upkick [OPTIONS]
Application Options:
-V, --version Display version.
-l, --loglevel= Set loglevel ('debug', 'info', 'warn', 'error', 'fatal', 'panic'). (default: info) [$UPKICK_LOG_LEVEL]
-m, --manpage Output manpage.
-j, --json Log as JSON (to stderr). [$UPKICK_JSON_OUTPUT]
-w, --warn-only Only warn, do not kick out-of-date containers. [$UPKICK_WARN_ONLY]
-H, --hostname-from-rancher Retrieve hostname from Rancher metadata. [$CONPLICITY_HOSTNAME_FROM_RANCHER]
Docker Options:
-e, --docker-endpoint= The Docker endpoint. (default: unix:///var/run/docker.sock) [$DOCKER_ENDPOINT]
Metrics Options:
-g, --gateway-url= The prometheus push gateway URL to use. [$PUSHGATEWAY_URL]
Help Options:
-h, --help Show this help message
Using the Docker image
$ docker run -v /var/run/docker.sock:/var/run/docker.sock:ro --rm -ti camptocamp/upkick
Per container opt-out
You can set containers to only warn if they are outdated by placing an io.upkick.warn_only=true
label on them.
Per container opt-in
If the global --warn-only
flag is passed, you can opt-in for container kicking by placing an io.upkick.warn_only=false
label on them.
Metrics
Upkick can push metrics to a Prometheus gateway. The currently exported metrics are:
upkick_containers{what="total",image="<image>"}
: total number of containers using a given image tagupkick_containers{what="blacklisted_tag",image="<image>"}
: number of containers using a given image that is blacklistedupkick_containers{what="blacklisted_container",image="<image>"}
: number of containers that opted-out of update (using labels)upkick_containers{what="up_to_date",image="<image>"}
: number of containers using a given image already up-to-dateupkick_containers{what="updated",image="<image>"}
: number of containers using a given image successfully updatedupkick_containers{what="update_failed",image="<image>"}
: number of containers using a given image whose update failedupkick_containers{what="not_updated",image="<image>"}
: number of containers using a given image that were not updated (because--warn-only
was used)
Here are some examples of useful Prometheus queries using these metrics: