Nearly sure it is not intended to delete helmreleases.helm.fluxcd.io
HelmReleases when users run `helm uninstall`, but the effect is there
currently, so until the behavior change needed is made clear, a more
strongly worded warning might be appropriate such as this. (See #811)
Is there any way this could be fixed? I would expect `flux uninstall` to
only remove resources that come from Flux v2 CRDs, not Helm Operator HRs
- add singular alias to get commands
- allow filtering the get commands result by resource name
- add the image commands to mkdocs index
Signed-off-by: Stefan Prodan <stefan.prodan@gmail.com>
ImageRepository objects can now refer to a secret containing
certificates to use for TLS. This adds the flag
flux create image repository --cert-secret-ref
for naming a secret to use. You can create such a secret with
flux create secret tls
Signed-off-by: Michael Bridgen <michael@weave.works>
This fixes a bug where the wrong type was displayed for various
`get source` commands.
```console
$ flux get sources helm --namespace default
✗ no Bucket objects found in default namespace
```
Signed-off-by: Hidde Beydals <hello@hidde.co>
The image-reflector controller now accepts a secret containing a
client certificate and key, and/or a CA certificate; so it's useful to
have a command for creating them.
`flux create secret helm` is close, but accepts username/password
(which would be ignored), and has the wrong name of course. Happily
though, much can be shared between the implementations.
Signed-off-by: Michael Bridgen <michael@weave.works>
* Use `LocalObjectReference` and `NamespacedObjectKindReference`
from `meta` package, as required by controller API changes.
* Remove `Update` field from created `ImageUpdateAutomation`,
as the API changed and the default is now defined in the Custom
Resource Definition.
Signed-off-by: Hidde Beydals <hello@hidde.co>
This commit deprecates the architecture flag (`--arch`) for the install
and bootstrap commands, in favor of the bundled multi-arch images that
will be available for the next MINOR range of GOTK controller releases.
Summary of changes:
* `*Arch` variables have been marked as deprecated for both commands.
* `-arm64` suffix is no longer selectively added to the image definition
of a component's `Deployment`.
* `kubernetes.io/arch` node selector with the defined value has been
removed from the components' `Deployment`s.
* `Arch` has been removed from the available `Options` in
`manifestgen/install`.
* Documentation references have been changed to highlight existence
of multi-arch images and supported architectures.
Signed-off-by: Hidde Beydals <hello@hidde.co>
controller-runtime methods now accept `client.Object` and
`client.ObjectList` rather than `runtime.Object`. This means the
adapter interfaces need to change signature, but happily, little else.
Since the list adapter is now distinct to the object adapter, `len()`
can go there instead of the command-specific interfaces.
Signed-off-by: Michael Bridgen <michael@weave.works>
This commit adds multiple safe guards for relative paths, ensuring they
never traverse outside the working directory.
The `SafeRelativePath` flag calculates the safe relative path based on a
relative base dir, which results in a flattened path.
The write methods of `manifestgen` make use of the `SecureJoin` as well,
to ensure writes are never outside of the given directory when used as
a lib outside of the CLI.
Signed-off-by: Hidde Beydals <hello@hidde.co>
It's a common pattern in the create commands to construct a value,
then (if not exporting it) upsert it and wait for it to
reconcile. This commit factors `upsert`, which does the update/insert
bit, and `upsertAndWait`, which does the whole thing.
Since these output messages, they are methods of `apiType` (previously
`names`), so that they have access to the name of the kind they are
operating on.
Signed-off-by: Michael Bridgen <michael@weave.works>
Kingdon Barrett
Stefan Prodan
Michael Bridgen
jonathan-innis
Somtochi Onyekwere
Hidde Beydals
Philip Laine
Alexei Ledenev
Léopold Jacquot