This commit moves the notification dispatching for creating composes and
clones into a more sensible location – the Image Builder API slice.
It is more sensible because it separates the logic of the React
component (the wizard or share images modal) from the logic of handling
the request life cycle (which is now handled entirely in the slice).
There is a subtle but significant change – a new request will be
dispatched for every request. This is the correct way to do things as it
is possible that some requests succeed, and that others fail. Insights
causes the notifications to stack on top of each other neatly, so there
is no UI problem.
To facilitate this, we also need to use use Promise.allSettled instead
of Promise.all.
There was a bug preventing all the partitions required by an oscap
policy to show up in the fsc step. That bug was because the FSC
customizations were only accepted if they matched exactly the fix set of
base partitions directories.
So now there's a bit of string manipulation to avoid that but it's all
patched up.
fixes https://github.com/RedHatInsights/image-builder-frontend/issues/1385
The images table uses the useGetComposesQuery hook to fetch composes and
implement pagination. When this query is used, args are provided for the
offset and limit and a 'Compose' tag is provided for the query.
When a mutation is triggered (causing a POST request to be sent to the
`/compose` end point), the 'Compose' tag is invalidated which clears all
cached data from useGetComposesQuery hooks, which in turn causes the
table to refetch compose information.
If invalidating the `Compose` tag causes a refetch before the new compose
is available in image-builder, the result does not contain the new
compose and the table is not updated to include it.
This commit eliminates the race condition by waiting for the query to be
fulfilled before invalidating tags (and therefore before refetching the
data).
All of the above applies equally to the `cloneCompose` mutation, and its
race condition has also been eliminated.
This commit is loosely inspired by the RTK Query docs section on
pessimistic updates:
https://redux-toolkit.js.org/rtk-query/usage/manual-cache-updates#pessimistic-updates
Typescript complains about the type of the tags. It does not recognize
the tag types that are defined in the same enhanceEndpoint() function.
For now, we simply ignore the Typescript errors. There is some
discussion here: https://github.com/reduxjs/redux-toolkit/issues/1510
When the user clicks on the "create an image" button, the image wizard
shows up replacing the landing page. This allows to keep the quickstart
guides to the right of the wizard while the user is interacting with it.
The compliance step is now tested with two new tests:
- One to test that upon creation, if a user selects a profile, then the
customizations are applied to the file system and package steps
- One to test that upon recreation, the file system and packages
customizations are still present.
HMS-2077
When the user has selected a profile to customize their distribution,
the FSC step now shows immediately the manual configuration opened up
with the partitions coming from the profile already filled in.
If the user goes back to the Compliance step and changes anything, the
FSC step is going to be reinitialized.
HMS-2077
Do not show the oscap step outside of beta. Only protecting the step
itself is sufficient since the FSC and the Packages steps wont load any
profile data if no entry point allows to get one.
HMS-2077
wizardrecreatebeta
This commit adds the Compliance step to the wizard. In this step the
user can select a policy from the list of available policies. Their
image is going to be updated with the necessary changes on the later
steps.
IB and The compliance endpoint are both returning the list of policies a
user has access to. The oscap step computes the intersection of the
policies accessible before showing the select list to the user.
HMS-2077
This reverts commit 36f9e70565.
The Compliance v1 API is not stable and not intended for consumption by
other teams. We will implement the OpenSCAP MVP without the Compliance
API for now with a hard coded list of polices in image-builder, and when
their v2 API is ready in the near future begin using it to retrieve
policies.
This commit adds the Compliance service's API, which we can query for a
list of SCAP security policies.
In a follow-up PR, we will allow the user to optionally select a policy
when creating an image in the wizard.
Adding GCP image sharing option according to the discussion on slack(https://redhat-internal.slack.com/archives/C03AZ0264LW/p1692789579814619) and mocks(https://issues.redhat.com/browse/HMS-2352).
In summary, within our GCP sources, we store the project ID. Images cannot be shared directly with a project ID, but they can be shared with a service account or a Google account. Consequently, to launch instances in GCP, users are not required to provide their Google account; instead, the images should be shared with the provisioning service account. This ensures that the provisioning team has access to the necessary images, as sharing them with individual users would not allow that.
After a thorough discussion, we have collectively decided to introduce an option. This option allows users to exclusively utilize the Launch service without the need to share the image with a Google account.
The `<LaunchWizard>` architecture prop was specified incorrectly,
which caused the Instance dropdown in the Launch wizard to fail to load
instance types.
This commit is a hotfix that restores functionality to the Launch
wizard.
The `appendTo` variable can't be recomputed on every render. Wrap it in
a useState() hook to work properly.
Additionally, the `hasAccess` property was missing from the
<ProvisioningWizard> component.
To ensure full backwards compatibility, the component has also been
wrapped in a <Suspense> as it was previously.
The edge images table is going to be integrated in the frontend code
base in the coming weeks. To help smooth that development process there
is need for a feature flag that would keep the current behavior running
on stage and prod while the new image table is being developed.
For now dummy code is returned for when the flag is on, and legacy
behavior keeps on going when the flag is off.
The stage flag can be seen there:
- https://insights-stage.unleash.devshift.net/projects/default/features/image-builder.edge.local-image-table
The prod flag can be seen there:
- https://insights.unleash.devshift.net/projects/default/features/image-builder.edge.local-image-table
The commit brings in some documentation in the Readme. Any member of
the team should be able to follow these steps to add in new feature
flags and use them in the code base.
The commit also brings in two new testing files for the future
components to come. For now these tests only check that the mocking
value for the unleash flag is properly set.
This commit converts the Images Table to Typescript and converts all API
calls to image-builder to use RTK Query hooks.
This should increase the performance of the app significantly.
Previously our calls to the image-builder API were made in series. They
are now made in parallel. We may want to investigate the possibility of
hitting rate limiting now that we will be issuing requests in much more
rapid succession.
In the tests, moving to RTK Query hooks has allowed us to remove
virtually all Jest mocking. However, this means that some of our
previous tests which tested against implementation details were broken.
Most notably, we no longer check the Redux store to verify that clones
have been added correctly and we no longer check that compose requests
were issued successfully. Test coverage will be restored in a follow-up
PR where the dev-dependency @msw/data is added. Adding a persistent data
layer to the tests using @msw/data will allow us to verify that our POST
requests (creating composes and cloning them) are working by testing
that the Images Table has been updated.
This updates the Repositories and Review step as per [mocks](https://www.sketch.com/s/d7aa6d29-fca0-4283-a846-09cc5fd10612/a/MyEbDz7).
Repositories with the unavailable or invalid status have a popover that allows for further inspection. The time of the last introspection and the counter of failed attempts was added to the popover, together with the "Go to Repositories" button.
On Recreate the payload repositories are checked against "freshly" fetched list of repositories. In case any of the previously checked repositories is no longer available in content sources an Alert is rendered on both Repositories and Review steps. The unavailable repository is checked, but the checkbox is disabled and the information is dashed out. Since the information about the repository is stored in the Repository type, the only information available to be rendered is the baseurl.
Create image button is also disabled when recreating an image with unavailable repositories.
The "Step Image output - clicking Next loads Upload to AWS" and "Step Upload to Azure - clicking Next loads Registration" were sometimes returning the "Warning: An update to <element> inside a test was not wrapped in act(...)" error.
This seems to have been caused by an early exit in the `setUp()` function in the first `describe` block.
Safeguards were put in place to fix this behaviour.
As this error didn't occur every time, testing the updates was not easy. This current fix passed the threshold of ten successful runs of `npm run test`.
After bumping to React 18 linter was outputting following error: "ReactDOM.render is deprecated since React 18.0.0, use createRoot instead, see https://reactjs.org/link/switch-to-createroot"
This fixes the problem.
Test `component renders error state correctly` was outputting errors for both AWS and Azure tests.
This was caused by calling `server.use()` after the wizard was already rendered.
Fixed by moving the one time server mock response to the very beginning of the unit test. Also added `server.resetHandlers()` to be run after each of the unit tests.
Thanks to this fix, azure tests could be merged again. Which means no more splitting into `CreateImageWizard.azure.test.js` and `CreateImageWizard.azure.2.test.js`
This updates following test suites after bumping react and testing-library:
- CreateImageWizard.azure.test.js
- CreateImageWizard.azure.2.test.js
- CreateImageWizard.content.test.js
- CreateImageWizard.test.js
Fixes HMS-1492
This changes all occurences of "rhc" and "remote host configuration" from capitalized to lowercase.
In cases where the sentence started with "rhc" this was changed to "remote host configuration" to avoid
starting the sentence with "Rhc".
Previously, the Launch/Provisioning wizard was only available in
preview. If an AWS or Azure image was created in preview and shared with
a source, and then viewed in the images table in stable, a popover over
the stable launch link would inform the user that the image had been
created in preview and provide a link to go to preview.
Now that launch/provisioning is going GA and the launch wizard is
available in stable, this feature is no longer necessary.
