3 ms·

What tool do you use to manage Kubernetes manifests? Do you not have any issues when needing to manage tens, hundreds, or thousands of different deployments ac

What tool do you use to manage Kubernetes manifests? Do you not have any issues when needing to manage tens, hundreds, or thousands of different deployments across multiple Kubernetes clusters?

mplewis 3y ago
Pulumi solves these problems for me, and it’s type-safe too.
- yowlingcat 3y ago
  We aren't on K8S, but chose Pulumi specifically because it gave a unified way to manage deployment manifests in code across things that are currently in AWS/managed services (all of it, for now), and things we might later put into K8S. We've been very happy with it.
prmoustache 3y ago
All continuous deployment tools i have used like flux and argosCD can deploy YAML/json manifests stored in plain directories of a git repo. Kustomize allows to manage different environments/configuration options easily.
- xyzzy_plugh 3y ago
  Kustomize is pretty horrible stuff. I wouldn't want to use it for anything serious.
jml78 3y ago
I hate helm. For us, it led to a PITA knowing where values came from. You had to debug logic in the helm charts to know why a value existed in the cluster with no good debugging tools. We switched to kustomize with tooling around it. It is dead simple to know why a value landed in an environment this way at the expense of yaml proliferation but I would rather have tooling to help us manage that than trying to debug why a value got into production due to an bug in helm logic in the templates. I don’t want to debug service logic and helm templating logic.
- mdaniel 3y ago
  > I hate helm. For us, it led to a PITA knowing where values came from Do you mean "came from" like "who is the rando that plugged in rm -rf" or "came from" like how did this Deployment end up with replicas:2 ? If the latter, $(helm get values) <https://helm.sh/docs/helm/helm_get_values/ https://helm.sh/docs/helm/helm_get_values/> goes a long way toward that end

What tool do you use to manage Kubernetes manifests? Do you not have any issues when needing to manage tens, hundreds, or thousands of different deployments across multiple Kubernetes clusters?

mplewis 3y ago

Pulumi solves these problems for me, and it’s type-safe too.

yowlingcat 3y ago
We aren't on K8S, but chose Pulumi specifically because it gave a unified way to manage deployment manifests in code across things that are currently in AWS/managed services (all of it, for now), and things we might later put into K8S. We've been very happy with it.

prmoustache 3y ago

All continuous deployment tools i have used like flux and argosCD can deploy YAML/json manifests stored in plain directories of a git repo. Kustomize allows to manage different environments/configuration options easily.

xyzzy_plugh 3y ago
Kustomize is pretty horrible stuff. I wouldn't want to use it for anything serious.

jml78 3y ago

I hate helm. For us, it led to a PITA knowing where values came from. You had to debug logic in the helm charts to know why a value existed in the cluster with no good debugging tools. We switched to kustomize with tooling around it. It is dead simple to know why a value landed in an environment this way at the expense of yaml proliferation but I would rather have tooling to help us manage that than trying to debug why a value got into production due to an bug in helm logic in the templates. I don’t want to debug service logic and helm templating logic.

mdaniel 3y ago
> I hate helm. For us, it led to a PITA knowing where values came from Do you mean "came from" like "who is the rando that plugged in rm -rf" or "came from" like how did this Deployment end up with replicas:2 ? If the latter, $(helm get values) <https://helm.sh/docs/helm/helm_get_values/ https://helm.sh/docs/helm/helm_get_values/> goes a long way toward that end