#helm - September 2019

tricky question

but I'm sure you peeps have an answer already...

what if I have legacy k8s artifacts.. (deployments, services, cronjobs)

but I want to import them into helm state

use case: migrate old circle-ci workflow -> new ci/cd tool using helm/helmfile

(old circle-ci workflow would just update docker image in place for all the artifacts)

i'm trying not to create new deployments/services since that would create new ALBs/Route53

i mean.. if I have to, I will

.. but just trying to make a seamless transition

use raw chart?

+ helmfile

lemme see what raw does

lol 😅

We have a bunch of examples here

Hmm, I mean, i'm less concerned about making the helm charts/helmfiles. I just kubectl get deployment blah -o yaml and created charts for everything

it's just if I do a helmfile apply and the deployment name of blah is already taken, helm won't deploy it.. and not keep track of it in helm state

raw chart is super interesting though

still looking at it

use yq or jq to strip the namespace from the kubectl get export

you'll also want to strip a few other things when exporting the resources

i guess i'm struggling with seeing the part you're concerned about 😃

1. exporting resources with kubectl get - easy

2. stripping metadata like namespace - easy

3. using the raw chart with helmfile - easy

ohhh

sounds like you want to do more than just import the existing state

sounds like you want to parameterize the Deployment name too

that's a slippery slope. at that point, you're not really "importing" the existing state which the raw chart would help you do.

hopefully deploying to multiple namespaces is sufficiently parametetrized

if you need to do more than that, i would either (a) write a helm chart (b) use the monochart

have you seen the monochart?

Yeah, monochart is awesome. But yeah, I'm totally not worried about creating helm charts or helmfiles.

It's just when I run helmfile apply it fails because it says deployment hello-world-dev already exists

this is more a long the lines of my concern

Aha ya, guess that part may be messy

I see there are proposals for import like terraform import, but yeah

heh, no worries

curious if I can just rebuild the state and pop in the ConfigMap.. and register it somehow.. then have it detected by helm ls

https://github.com/helm/helm/blob/master/_proto/hapi/release/status.proto#L31
hhmmmm

kubectl -n kube-system get cm hello-world.v1  -o yaml | grep release | cut -d' ' -f4 | base64 -D | gunzip

I think the first section is protobuf.. just gotta flip the right byte...

https://gist.github.com/rms1000watt/3d76eb2f3c64a87b92acb97ebdbd9c66 helm dark arts.. helm ls now says DEPLOYED.. lets see if I can deploy again without failure

sweet.. helmfile apply works fine now

wow, you went deep down that rabbit hole

that's a cool trick