Runs `helm version` in helmexec.New, and exposes a method on Interface to allow other packages to use the detected version. Preserves compatibility with previous HELMFILE_HELM3 mechanism.
Resolves#923
* Fix regression since 0.90.0 that Helmfile becomes too slow when there are many releases
Fixes#959
* Ensure that the up-to-date helm-diff is installed and used in integration tests
The problem was that `--namespace NS` had been not taken into account while deleting releases, that resulted in releases that should be deleted are not deleted.
The recent addition of the DAG support(`needs`) and the fixes on it broke the delete-on-sync functionality. And there were two more bugs. One is that it was not correctly running `helm delete` when needed and the another is that it was failing when `--selector` is specified and the releases to delete by sync found, but nothing actually got deleted. This fixes all of them.
Fixes#941
This ports the fix for `helfmile apply` to `sync`, so that specifying `--selector` doesn't break `helmfile sync`.
Also make `helmfile template` DAG-aware, so that the manifests are rendered in the order of dependency.
Ref #919
`helmfile delete` is designed to be idempotent. That is, it is safe to be run when there is nothing to delete. A change in helm v3 broke that behavior. This enhances Helmfile to be able to behave the same for helm v2 and v3.
Until now `helmfile delete` with helm v3 had been failing when `--namespace` is provided. Helmfile was missing the fact that in helm v3 namespace must be passed instead of tiller-namespace to scope releases to be deleted.
```
err: release "b-1" failed: helm3rc2 exited with status 1:
Error: uninstall: Release not loaded: b-1: release: not found
in ./helmfile.yaml: release "b-1" failed: helm3rc2 exited with status 1:
Error: uninstall: Release not loaded: b-1: release: not found
```
This overrides the default helm command, if provided, as soon as possible.
This way it is already used in `visitStates`.
I ran into an issue using `HELM3` and `--helm-binary` together with helm-secrets.
I previously used tillerless, which i could now remove. This however caused `DecryptSecret` to fail, as it would still use the helm2 binary; because it runs before the first `helm.SetHelmBinary` call.
While helm-secrets is not fully helm3 compatible yet, its uses within helmfile are, and i was able to just install it as a helm3 plugin.
* fix: Fix `needs` to work for upgrades and when selectors are provided
Fixes#919
* Add test framework for `helmfile apply`
* Various enhancements and fixes to the DAG support
- Make the order of upgrades/deletes more deterministic for testability
- Fix the test framework so that we can validate log outputs and errors
- Add more test cases for `helmfile apply`, along with bug fixes.
- Make sure it fails with an intuitive error when you have non-existent releases referenced from witin "needs"
Introduces DAG-aware installation/deletion ordering to Helmfile.
`needs` controls the order of the installation/deletion of the release:
```yaml
relesaes:
- name: somerelease
needs:
- [TILLER_NAMESPACE/][NAMESPACE/]anotherelease
```
All the releases listed under `needs` are installed before(or deleted after) the release itself.
For the following example, `helmfile [sync|apply]` installs releases in this order:
1. logging
2. servicemesh
3. myapp1 and myapp2
```yaml
- name: myapp1
chart: charts/myapp
needs:
- servicemesh
- logging
- name: myapp2
chart: charts/myapp
needs:
- servicemesh
- logging
- name: servicemesh
chart: charts/istio
needs:
- logging
- name: logging
chart: charts/fluentd
```
Note that all the releases in a same group is installed concurrently. That is, myapp1 and myapp2 are installed concurrently.
On `helmdile [delete|destroy]`, deleations happen in the reverse order.
That is, `myapp1` and `myapp2` are deleted first, then `servicemesh`, and finally `logging`.
Resolves#715
At the moment, if you have a helmfile.yaml like so:
```
releases:
- name: metrics-server
namespace: kube-system
chart: stable/metrics-server
```
If you try to run `helmfile deps`, you will get a 0 exit code and no log
output at whatsoever, signaling that there weren't any problems, but no
lock file will get created.
For example:
```
root@316073d4a104:/# helmfile deps
root@316073d4a104:/#
```
This behavior doesn't appear to be documented and is unintuitive to the
user.
This change adds a warning output for this same use case:
```
root@316073d4a104:/# helmfile deps
There are no repositories defined in your helmfile.yaml.
This means helmfile cannot update your dependencies or create a lock file.
See https://github.com/roboll/helmfile/issues/878 for more information.
root@316073d4a104:/#
```
Fixes#878
Set `HELMFILE_HELM3=1` and run `helmfile` like `HELMFILE_HELM3=1 helmfile ...`.
When `HELMFILE_HELM3` is set, `test`, `template`, `delete`, `destroy` behave differently so that it works with Helm 3.
Note that `helmfile diff` doesn't work as `helm-diff` called under the hood doesn't support Helm v3 yet.
Ref #668
Closes#444 and #782
This is the final PR to fully cache and parallelize helm secret decryption. It threads the shared helmexec.Interface into the StateCreator and HelmState structs to be used during environment secret decryption. This should effectively cache secrets for the duration of a helmfile run, regardless of where they are first decrypted.
Related to #782 and #444
- Allows concurrent decryption of different secrets files
- Caches decrypted secrets by original file path and returns decrypted results from memory
- Secrets being run through an instance of helmexec will be cached and run as fast as possible concurrently
NB: This particular PR doesn't make _all_ calls to secrets cached and concurrent. Environment Secrets in particular seem to not be evaluated with a ScatterGather(), and doesn't use the same helmexec instance as other parts of the code, so it doesn't take advantage of these changes. Some reworking of the plumbing there would be needed.
Use with the helm-x support(#673)
This enhances config syntax to accept adopt: [NS/KIND/RESOURCE_NAME] at the release level so that helmfile calls helm-x to transparently import existing resources at the installation time.
Resolves#84
Resolves#689
This adds a new yaml entry for the hook definition to allow the users to specifcy if they want to show the `command` logs or not.
here is an example.
```
releases:
- name: myapp
chart: mychart
# *snip*
hooks:
- events: ["cleanup"]
showlogs: true
command: "kubectl"
args: ["get", "ingress"]
```
this will display the following output:
```
hook[cleanup] logs | NAME HOSTS ADDRESS PORTS AGE
hook[cleanup] logs | catalog-gateway tdc.foo 80 2d6h
hook[cleanup] logs | dataset foobar.barr.foo.xxxxxxx.com 80 2d6h
hook[cleanup] logs | rating fooba.barr.foo.xxxxxxx.com 80 2d6h
hook[cleanup] logs | sharing foobar.barr.foo.xxxxxxx.com 80 2d6h
hook[cleanup] logs | tpsvc-iam-dev foo.barr.foo.xxxxxxx.com 80 2d6h
hook[cleanup] logs | tpsvc-iam-front bar.barr.foo.xxxxxxx.com 80 2d6h
```
The root cause of this bug was due to that `--kube-context` and `kubeContext` had been treated specifically in code. So on the way I have made it consistent with other per-release settings - by adding `kubeContext` for each release and treating `helmDefaults.kubeContext` as just the default value for per-release setting.
Fixes#674
This enhances helmfile so that it can:
- Treat K8s manifests directories and Kustomize projects as charts
- Add adhoc chart dependencies on sync/diff/template without forking or modifying chart(s) (#649)
- Add adhoc patches(JSON Patch or Strategic Merge Patch supported) to be applied to the K8s resources before sync/diff/template, without forking or modifyin chart(s) (#650)
The usage is as outlined in https://github.com/mumoshu/helm-x/tree/master/examples/helmfile.
Add any or all of `dependencies:`, `jsonPatches:` and `strategicMergePatches:` so that it adds additional flags to `helm` calls that is only supported by `helm x`.
```yaml
releases:
- name: kustomize
chart: ../kustomize
- name: manifests
chart: ../manifests
- name: foo
chart: incubator/raw
dependencies:
- alias: bar
chart: incubator/raw
values:
- values.yaml
- bar:
enabled: true
resources:
- apiVersion: v1
kind: Pod
metadata:
name: bar
spec:
containers:
- command:
- sleep
- 1000
image: alpine:3.9.4
imagePullPolicy: IfNotPresent
name: bar
jsonPatches:
- target:
version: v1
kind: Pod
name: foo
patch:
- op: replace
path: /spec/containers/0/command
value:
- sleep
- "123"
strategicMergePatches:
- apiVersion: v1
kind: Pod
metadata:
name: bar
spec:
containers:
- name: bar
command:
- sleep
- "234"
```
You can alternatively provide `source: path/to/patch.yaml` for `jsonPatches` and `strategicMergePatches` items to externalize it. Add `.gotmpl` suffix like you would do for values files for templating.
When running `helmfile` you must point `--helm-binary` to the `helm-x` binary like below:
```
$ helmfile --helm-binary ~/.helm/plugins/helm-x/bin/helm-x --log-level debug apply
```
after installing the [helm-x](https://github.com/mumoshu/helm-x) plugin.
The integration should ideally be automatic. That is, it shouldn't force you to set `--helm-binary`. But I had no other way to not bloat helmfile's codebase to just add this experimental feature.
Resolves#649Resolves#650
Probably since #647 helmfile has been unable to merge nested maps in environment values if they were loaded from files. This fixes it.
The relevant test is also enhanced so that no further regression like this happens.
Fixes#677
Extends the remote-helmfile feature to also work when loading the first state file.
This should be useful for people who wants to give helmfile a try without ever opening `$EDITOR`.
* fix: persist original file path when using bases
Prior to this change, the resulting lock file was called `<bases[0]>.lock`,
instead of `<filename>.lock`.
This change ensures the final, merged state has the correct `.FilePath`.
* test: Assert proper FilePath in layered HelmState
helm-secrets uses the `HELM_SECRETS_DEC_SUFFIX` env var to define the name of the output file
we should have the same logic in helmfile, to come up with the same filename
It only affects people using the `HELM_SECRETS_DEC_SUFFIX` env var
Use-case: if you want to run multiple `helmfile` commands in parallel, without conflicts. in this case, you need to decrypt secrets with different suffixes.
This change enhances helmfile to accept terraform-module-like URLs in nested state files a.k.a sub-helmfiles.
```yaml
helmfiles:
- # Terraform-module-like URL for importing a remote directory and use a file in it as a nested-state file
# The nested-state file is locally checked-out along with the remote directory containing it.
# Therefore all the local paths in the file are resolved relative to the file
path: git::https://github.com/cloudposse/helmfiles.git@releases/kiam.yaml?ref=0.40.0
```
The URL isn't equivalent to terraform module sources. The difference is that we use `@` to distinguish between (1) the path to the repository and directory containing the state file and (2) the path to the state file being loaded. This distinction provides us enough fleibiity to instruct helmfile to check-out necessary and sufficient directory to make the state file works.
Under the hood, it uses [hashicorp/go-getter](https://github.com/hashicorp/go-getter), that is used for [terraform module sources](https://www.terraform.io/docs/modules/sources.html) as well.
Only the git provider without authentication like git-credentials helper is tested. But theoretically any go-getter providers should work. Please feel free to test the provider of your choice and contribute documentation or instruction to use it :)
Resolves#347
This adds `values` to state files as proposed in #640.
```yaml
values:
- key1: val1
- defaults.yaml
environments:
default:
- values:
- environments/default.yaml
production:
- values:
- environments/production.yaml
```
`{{ .Valuese.key1 }}` evaluates to `val1` if and only if it is not overrode via the production or the default env, or command-line args.
Resolves#640
Seems like we are affected by https://github.com/golang/go/issues/24963. That is, even though we internally use the template option `missingkey=zero`, in some cases it still prints `<no value>` instead of zero values, which has been confusing the state yaml parsing.
This fixes the issue by naively replacing all the remaining occurrences of `<no value>` in the rendered text, while printing debug logs to ease debugging in the future when there is unexpected side-effects introduced by this native method.
Fixes#553
The addition of `--set k1=v1,k2=v2` and `--values file1 --values file2` was originally planned in #361.
But it turned out we already had `--values` for existing helmfile commands like `sync`. Duplicated flags doesn't work, obviously.
So this actually add `--state-values-set k1=v1,k2=v2` and `--set-values-file file1 --set-values-file file2`.
They are called "state" values according to the discussion we had at #640Resolves#361
* feat: helmfile as a go library
This removes almost all the dependencies from the helmfile core logic to urfave/cli. `main.go` is now a thin wrapper around the core logic implemented in `pkg/app`.
We added envvals overrides in the state file via #622 two days ago:
```
helmfiles:
- name: sub.helmfile.yaml
environment:
values:
- mykey: myvalue
```
This change removes the `environment` level in the above cofig, so that it looks like:
```
helmfiles:
- name: sub.helmfile.yaml
values:
- mykey: myvalue
``
This is an inevitable breaking change towards #361. But I wanted to break it earlier so that less folks are affected.`
Ref https://github.com/roboll/helmfile/issues/361#issuecomment-497530819
* feat: specify env values from the parent to the nested state
Adds the `helmfiles[].environment.values` that accepts a mix of file pathes and inline dictes:
```yaml
helmfiles:
- path: path/to/nested/helmfile.yaml
environment:
values:
- key1: val1
- values.yaml
```
The values files are loaded in the context of the parent state file. For example, in case the above state file is located at `/path/to/helmfile.yaml`,
`values.yaml` is located at `/path/to/values.yaml` instead of `/path/to/nested/values.yaml`.
Resolves#523
* fix: multiple "bases" declarations yields duplicate releases
Fixes#615
* fix regression in double-rendering with env value overrides
The latest commit broke any state files like the below to NOT pass env value overrides at all:
```
helmfiles:
- path: nested/state.yaml
environment:
values:
- overrides.yaml
```
This fixes the issue.
```yaml
environments:
default:
missingFileHandler: Warn
values:
- path/to/values.yaml
secrets:
- path/to/secrets.yaml
```
`missingFileHandler` set to `Warn`, `Info`, or `Debug` results in helmfile NOT stop when `path/to/values.yaml` or `path/to/secrets.yaml` is missing.
Resolves#548
While implementing the above feature, I also found a bug that has been causing #559. This also fixes that.
To verify it is actually fixed, create an example helmfile.yaml that looks like the below, and run `helmfile diff`:
```
$ cat helmfile.yaml
environments:
default:
secrets:
- env-secrets.yaml
releases:
- name: myapp
chart: nginx
namespace: default
secrets: [secrets.yaml] # Notice this file does not exist
values:
- ingress:
enabled: true
$ helmfile diff
could not deduce `environment:` block, configuring only .Environment.Name. error: failed to read helmfile.yaml.part.0: environment values file matching "env-secrets.yaml" does not exist
in ./helmfile.yaml: failed to read helmfile.yaml: environment values file matching "env-secrets.yaml" does not exist
```
Fixes#559
This enhances helmfile's internal environment values files loader to expand glob patterns (#606)
Fixes the existing bug that helmfile was unable to load environment values file from absolute path (#549)
Resolves#606Fixes#549
* Various U/X improvements for `helmfile apply`
This improves the U/X of `helmfile apply`, by allowing you to selectively apply sub-helmfiles.
When you have two or more sub-helmfiles processed, typing `n` to cancel the first doesn't automatically stop the whole helmfile execution.
Instead, it proceeds by diffing the next sub-helmfile, and asks you to apply it, which should be what the user would expect.
To support this workflow, I have suppressed useless exec logs, correct exit status when diff exists in sub-helmfiles but not in the parent helmfile, and made the final error message emitted by helmfile better.
More concretely, this moves more output from `helm` to STDERR and the `debug` log-level.
The overall output from `helmfile` should be a bit more cleaner especially for `apply`, `sync`, `diff` and perhaps other `helmfile` sub-commands, too.
For example, when one of release failed, `helmfile`'s final error message now includes the error message from the failed `helm` execution, like seen in the last line:
```
List of updated releases :
RELEASE CHART VERSION
envoy stable/envoy 1.5.0
List of releases in error :
RELEASE
envoy2
in ./helmfile.yaml: in .helmfiles[0]: in /Users/c-ykuoka/helmfile/helmfile.1.yaml: failed processing release envoy2: helm exited with status 1:
Error: UPGRADE FAILED: "envoy2" has no deployed releases
```
This way you can better understand what caused helmfile to finally fail.
`helmfile` has been streaminig a lot of stdout and stderr contents from the `helm` commands regardless of the helmfile's log-level. It has been suppressed by default and moved to the `debug` log-level.
You will see that it helps you focus on what was the cause of a failure.
While working on the above, I found an another bug that made `--detailed-exitcode` useless in some case.
That is, `helmfile diff --detailed-exitcode`, when any diff existed only in sub-helmfiles, has been returning an exit code of `1`. It should return `2` when any release had diff and no release had an error, regardless of the target is a sub-helmfile or a parent helmfile. Why? Because that's what `--detailed-exitcode` meant for!
After this PR gets merged, `helmfile diff --detailed-exitcode` propery return exit code `2` in such cases.
Fixes#543Resolves#540
Fixes#344 by allowing explicit selectors to be specified for composed helmfiles using the following structure
```yaml
helmfiles:
- path: helmfile.d/a*.yaml
selectors:
- name=prometheus
- name!=zipkin
- helmfile.d/b*.yaml
- path: helmfile.d/c*.yaml
selectors: {}
```
2 modes here :
* legacy mode when no the env var HELMFILE_EXPERIMENTAL is not set to true
* no selector : inherit from the command line.
* selector: is specified then it is used (an emty means no inheritance from command line and take everything).
* experimental when the env var HELMFILE_EXPERIMENTAL=true
* no selector : nothing is inherited from the command line so use all releases.
* selector: is specified then it is used (an emty means no inheritance from command line and take everything).
This basically allows to define both `releases` and `helmfiles` within a helmfile.yaml, so that you can start using sub-helmfiles easily, by extracting only reused releases.
Resolves#445
* Improve code organization
To make sure it is still readable after upcoming changes to helmfile
* feat: `helmfile deps` to update dependencies of all the local charts
Resolves#450
* feat: helmfile updates repos and build deps by default
But not update deps. Use `helmfile deps` to update deps, and provide `--skip-deps` to skip updating repos and builds deps in sync/diff/apply/template
Resolves#415
* Improve integration test coverage
Jeroen Demeyer
Andrew Drake
Andrew Drake
KUOKA Yusuke
Jan Möller
Tom Duffield
Kyrylo Lebediev
Jarryd Tilbrook
Aaron Batilo
Mike Splain
Theo Meneau
art kon
eddycharly
刘相轩
Shane Starcher
astorath
Travis Groth
OlivierB
a-hat
Marc Sutter
Erik Forsberg
sgandon
Eric Bailey
Kevin J. Qiu
Vincent Behar
Andrey Afoninsky
Patrick Valsecchi