[10:18:42] https://github.com/kubeshark/kubeshark [10:20:04] is that for local development? [10:20:25] (given the brew installation instructions ;) ) [16:08:44] how do folks test out their changes to operations/deployment-charts? [16:09:10] for example if I wanted to test out incorporting the jaeger helm chart, how would I test my patch? [16:09:22] is it possible to use operations/deployment-charts with minikube locally? [16:14:37] <_joe_> yes [16:14:40] <_joe_> also [16:14:55] <_joe_> you want to test the service or that the chart works? [16:16:23] test that the chart works [16:16:48] anyone have any notes on using operations/deployment-charts with minikube? [16:18:09] I have a working solution with helmfile, https://gitlab.wikimedia.org/repos/sre/jaeger-minikube, so now I want to incorporate the jaeger subste of that repo into operations/deployment-charts [16:18:19] *subset [16:43:30] <_joe_> jhathaway: I just run helm deploy [16:45:56] jhathaway: deployment-chars has CI btw [16:46:03] _joe_: hmm I'm confused, my helm version doesn't have a deploy command, do you mean the deploy.sh script in git root directory? [16:46:09] he meant apply [16:46:14] <_joe_> jhathaway: apply, sorry [16:46:26] <_joe_> I dropped a "the" [16:47:07] <_joe_> but, if you want to verify the chart is "correct" in k8s terms, CI takes care of it (mostly) [16:49:16] jhathaway: btw, you can also do helm repo add [16:49:17] see https://wikitech.wikimedia.org/wiki/ChartMuseum [16:49:32] for merged charts you can just follow that path [16:50:22] for local development of a chart, going to the directory of the chart and doing helm apply . --set value1=foo --set value2=bar etc works [16:53:02] hmm, okay that helps, still not totally grokking the workflow [16:57:29] why are the helmfile bits mixed into the same repo with helm charts? [17:01:09] these folks recommend a local instance of chartmuseum for development, do some folks do that? https://merative.github.io/spm-kubernetes/prereq/chartmuseum/ [21:24:37] jhathaway: monorepo approach. Single ci for both charts and deployments, easier git grep, testing, dev, etc. [21:26:51] Generic workflow for the end user services is create a new chart using the create new service bash script, iterate until you are happy, upload to gerrit, make ci happy, get reviewed, address, merge. Then a change in helmfile.d per deployment of that chart (e.g. eventgate chart has 4 deployments) [21:27:24] Make ci happy, get reviewed, merge, deploy [21:29:24] The cluster components aren't that different. The main diff would be that you can avoid using the bash script for creating the new chart, and that last I checked ci doesn't support admin_ng [21:31:01] As far as a local chart museum goes, I havent found the need for that. Same goes for a local oci image registry. No need for either (yet) [21:39:15] thanks akosiaris, so during the interation phase, are you testing somehow, or just hoping your yaml is correct? [21:41:49] I was hoping I could spin up aux-k8s-eqiad in minikube and then iterate on the chart locally until it was working correctly, is that the wrong approach? I see a smattering of references in the repo to minikube, but no realy guide on how to run pieces of the repo in minikube [22:03:48] jhathaway: I used https://wikitech.wikimedia.org/wiki/Local-charts when I was developing the chart for Toolhub. The tutorial at https://wikitech.wikimedia.org/wiki/Deployment_pipeline/Migration/Tutorial#Creating_a_Helm_Chart gives a general idea of how to do that. [22:05:13] bd808: thanks, I'll take a look at that [22:39:12] akosiaris: delayed \o :)