appnexus / ankh Goto Github PK

It looks like exactly one newline at the end of any matched document is removed, causing yaml badness when more than one document is matched and they are concatenated.

This is probably fallout from the highly hand-written yaml filter code that attempts to preserve comments and whitespace. More care is needed.

ankh config init should create ~/.ankh/config if it doesn't exist yet

fetch remote chart (as yaml?) and (print to stdout?)

after a fresh install, running ankh version prints this error:

$ ankh version
# WARNING /Users/ogupte/.ankh/config: Unable to read ankh config '/Users/ogupte/.ankh/config': open /Users/ogupte/.ankh/config: no such file or directory
# ERROR   Context '' not found in `contexts`
# INFO    The following contexts are available:

repro step
remove slack from your ankh config and try to execute ankh apply --slack, it will throw an error saying that authentication failed

expected:
apply validation if --slack is provided so that
if (sendSlackMessage && (!config.slack || !config.slack.token || !config.slack.format .... )) throw error

Steps to reproduce

Make a minimal-chart directory with the following Chart.yaml:

name: my-app
version: 1.0.0

And the following minimal-chart/templates/config-map.yaml:

apiVersion: v1
kind: ConfigMap
metadata:
  name: "my-app-{{ .Release.Name }}"
  labels:
	app: "my-app"
    release: {{ .Release.Name }}
data:
  SOME_CONFIG: "any-value"

Then run ankh lint chart-path=./minimal-chart, choose some environment, and leave the namespace blank.

Observed

Ankh panics with:

panic: attempted to parse unknown event: none [recovered]
	panic: attempted to parse unknown event: none [recovered]
	panic: attempted to parse unknown event: none

goroutine 1 [running]:
github.com/jawher/mow%2ecli.(*step).run(0xc000368600, 0x14e13a0, 0xc0002280f0)
	/Users/dbemiller/go/pkg/mod/github.com/jawher/[email protected]/flow.go:28 +0xd6
github.com/jawher/mow%2ecli.(*step).run(0xc000368660, 0x14e13a0, 0xc0002280f0)
	/Users/dbemiller/go/pkg/mod/github.com/jawher/[email protected]/flow.go:20 +0xbe
github.com/jawher/mow%2ecli.(*step).run(0xc0003691d0, 0x14e13a0, 0xc0002280f0)
	/Users/dbemiller/go/pkg/mod/github.com/jawher/[email protected]/flow.go:20 +0xbe
github.com/jawher/mow%2ecli.(*step).callDo.func1(0xc000369200, 0x0, 0x0)
	/Users/dbemiller/go/pkg/mod/github.com/jawher/[email protected]/flow.go:41 +0x4a
panic(0x14e13a0, 0xc0002280f0)
	/usr/local/Cellar/go/1.14.2_1/libexec/src/runtime/panic.go:969 +0x166
gopkg.in/yaml%2ev2.handleErr(0xc00073c798)
	/Users/dbemiller/go/pkg/mod/gopkg.in/[email protected]/yaml.go:249 +0x9a
panic(0x14e13a0, 0xc0002280f0)
	/usr/local/Cellar/go/1.14.2_1/libexec/src/runtime/panic.go:969 +0x166
gopkg.in/yaml%2ev2.(*parser).parse(0xc000356000, 0xc00007e100)
	/Users/dbemiller/go/pkg/mod/gopkg.in/[email protected]/decode.go:158 +0x251
gopkg.in/yaml%2ev2.(*Decoder).Decode(0xc00073c848, 0x14c7bc0, 0xc0001081e0, 0x0, 0x0)
	/Users/dbemiller/go/pkg/mod/gopkg.in/[email protected]/yaml.go:122 +0xdf
github.com/appnexus/ankh/helm.helmLint(0xc000362480, 0xc0000fe000, 0xbb, 0x0, 0x0, 0xc000600400)
	/Users/dbemiller/code/ankh/helm/lintstage.go:106 +0x232
...

Expected

An error message that describes what's broken or missing.

@esmet:

ankh passing templated output through temp files (instead of stdout/stdin pipes) would probably solve:

• ankh erroring immediately if the underlying kubectl wrapper prompts stdin for a password
• kubectl apply error messages to reference an input file line number instead of vaguely saying "from stdin"

Helm 3 removes the --name flag in favor of the following format:

Usage:
  helm template [NAME] [CHART] [flags]

This causes failures in ankh templating:

# INFO    Templating chart "flotilla" from path "helm/flotilla"
# FATAL   error running the helm command: exit status 1 -- the helm process had the following output on stderr:
Error: unknown flag: --name

Currently we print an error and tell the user to init but this should just be done automatically.

https://github.com/appnexus/ankh#chart
https://github.com/appnexus/ankh#chart-1
Is this intentional? If so it's confusing.

In any case what would be helpful to understand this section is to break down models by yaml file in which they can be found: from what I understand the ones at the top are in ~/.ankh/config and the ones at the bottom are in any ankh.yaml.

Using current-context is dangerous so let's remove it and instead allow people to use explicit --environment and --context flags.

maybe we can enumerate things to tidy up so that we can remove the "experimental" warning?

• perform ankh template
• perform ankh chart fetch (#87)
• print a diff of those two results

is it possible to use something like git diff --no-index for a nice +/- view (versus diff with </>)?

It's not currently possible to switch to a context that comes from an include:

$ ankh config use-context kube00nym2-staging
# ERROR   Context 'kube00nym2-staging' not found in `contexts`
# INFO    The following contexts are available:
# INFO    - minikube

We've decided that the whole concept of current-context should go away in the next major release since it's too dangerous to rely on that bit of state. In light of that, let's add a deprecation warning that says

WARNING current-context will be removed in the next major version of ankh. It currently does not support values that come from include. Use the --context flag instead.