Jenkins example: https://github.com/GoogleCloudPlatform/kubernetes/blob/master/contrib/recipes/docs/rollingupdates_from_jenkins.md

I haven't looked whether an off-the-shelf solution would largely work -- there certainly many workflow-y tools/systems out there.

Does this issue fall under the auspices of the continuous integration/delivery domain? e.g.:

• http://www.thoughtworks.com/continuous-delivery
• http://martinfowler.com/tags/continuous%20integration.html

At littleBits we are currently using https://circleci.com to glue a workflow that currently looks like:

local git -> Github -> CCI-Docker build ->

CCI-Docker test container -> CCI-Docker push to quay.io ->

CCI-kubekfg rolling update using image from quay.io

What's missing from above, and what we are currently working on, is how to get the basic information about:

• App version
• App branch
• App environment

wired to the deploy destination:

• What IP? FQDN? Port?
• Annotating the image with a tag that makes image-selection in the pod.yam config clear

Surely much of this is outside the realm of kubernetes (?) yet most/all of the above outlined seems foundational and basic to the needs of many (most?) micro-architecture systems?

CI/CD is more ambitious than what I was considering with this issue, which is more focused on building blocks for deploying updated configurations and/or images.

Indeed, CI/CD would be outside the realm of Kubernetes. Another example of adding such functionality to Kubernetes would be OpenShift.

@bgrant0607 Yeah 100% agree. Especially after reading https://github.com/openshift/openshift-pep/blob/master/openshift-pep-013-openshift-3.md I see better now what should/could be built on top of Kubernetes.

Typical scenario: Rollout across several clusters.

I'm trying to figure out how to change this subject line to properly reflect that it's covering deployment of Kube objects specifically. When I saw the subject I got excited because I thought there may have been discussion on a more robust tool to do cluster deployment and upgrade and replace kube-up.sh, which was an issue I was about to go write. These need disambiguation. :)

Disambiguated

I've written a tool for CI / CD using kubectl to facilitate most of this - be happy to get feedback or maybe this could serve as a POC for kubectl features https://github.com/UKHomeOffice/kb8or

Just a note that application deployment is usually the easy part. Client can usually retry until a service is available. A more difficult problem is application shutdown. These must usually be carefully orchestrated so as not to lose data. Data processors must be signaled to flush their data to the storage layer before they are terminate and the storage layer can't be terminated before all it's clients have finished their work.

Examples:
https://github.com/Mirantis/k8s-AppController
https://cloud.google.com/deployment-manager/configuration/create-configuration-file

https://github.com/atlassian/smith
Here is a prototype of a similar idea (if I understand this issue correctly). The idea came to me a while ago and then I found this and other similar issues. Hopefully it may be useful to someone for inspiration, etc. Any feedback is appreciated.

Closing as per discussion in #25067 (comment)