controller
controller copied to clipboard
deis
Adding lifecycle hooks to deis-controller api
Adding the lifecycle hooks feature to the deis-controller api. This allows users to attach lifecycle postStart and preStop hook commands to kubernetes api via the deis api.
Example of a preStop hook that would gracefully shutdown Resque worker processes: https://gist.github.com/Cryptophobia/447ecea7858d26141359b22161e70ee2
Example of the POST to deis.$hostname/v2/apps/testing-again/config:
JSON Body:
"lifecycle_pre_stop": {
"worker": "echo 'Running the postStop script' && chmod +x bin/kube-resque-pre-stop.sh
&& ./bin/kube-resque-pre-stop.sh;"
}
For long running preStop hooks the terminationGracePeriodSeconds needs to be increased for the worker pods:
deis config:set KUBERNETES_POD_TERMINATION_GRACE_PERIOD_SECONDS=900 -a testing-worker
The tests are not written yet. Code in deis-cli has not been written yet.
@rvadim : Yes, it would give you a graceful showdown option now. However, you still need to make sure that bash shell is present in the docker container that you are deploying. I was originally testing with one of the docker example Go apps and interestingly, the bash shell is not in there...
@mboersma : How do I get passed these three workflow-e2e failures in the build, they do not seem to be related to the deis-controller changes here?
Codecov Report
Merging #1311 into master will not change coverage. The diff coverage is
n/a.
@@ Coverage Diff @@
## master #1311 +/- ##
=======================================
Coverage 86.84% 86.84%
=======================================
Files 45 45
Lines 3944 3944
Branches 682 682
=======================================
Hits 3425 3425
Misses 348 348
Partials 171 171
The remote upstream/master was merged on top of the branch rather than rebase. Is that okay?
Is that okay?
The project has preferred git rebase just to keep merge commits at a minimum, but no worries--not a blocker.
Looks like Jenkins failed this randomly. I'll run it again and 🤞.
@mboersma : Looks like the build is still stuck in pending? It looks like it passed when you view the text output log.
How do I get this one passed as well? Looks like it failed because of a network issue related to ?missing namespace? in the Kubernetes cluster?
Jenkins, will you please pass this? Jenkins, no don't eat that off the floor! Jenkins, sit! Jenkins, be nice to me and I'll give you a cookie. 🍪 🐩
Looks like it failed because the Jenkins server could not create the "deis" user group. Seems like a server error. @mboersma , what do you think?
adduser: `/usr/sbin/groupadd -g 106 deis' returned error code 10. Exiting.
This PR has been recreated for the https://github.com/deisthree fork here: deisthree/controller#3