Added support for machine image update strategy

[grolu]

• Added support for machine image update strategy
• Use GMultiMessage component for Kubernetes version on GNewShootDetails
• Fixed empty tooltip on worker groups
• Show kubernetes version update information as list

What this PR does / why we need it:

Which issue(s) this PR fixes:
Fixes #1793

Special notes for your reviewer:

Release note:

The Dashboard now recognizes and displays automatic update notifications according to the configured `update strategy` for machine image vendors

[gardener-robot]

@holgerkoser You have pull request review open invite, please check

[holgerkoser]

This should not be logged as error and not inside the getter. I prosose that we remove it now.

[petersutter]

I agree that it should not be logged (as an error) inside the getter. However, if there is no place where we log that we skipped some entries, this might make it harder to troubleshoot issues where an entry is expected to be there but is not.

[grolu]

I think the main Problem here is that the machine image getter is not cached (because of param). That's why we get the log message multiple times. We should solve the root problem. @holgerkoser removed it in his refactoring PR afaik

[holgerkoser]

I have moved the flattening of the machineImages into an own function and call it right after data fetching. The getter now directly returns the already flattened machine images.

  async function fetchCloudProfiles () {
    const response = await api.getCloudProfiles()
    for (const cloudProfile of response.data) {
      set(cloudProfile, 'data.machineImages', flattenMachineImages(get(cloudProfile, 'data.machineImages')))
    }
    list.value = response.data
  }
  
  function flattenMachineImages (machineImages) {
    return flatMap(machineImages, machineImage => {
      const versions = filter(machineImage.versions, ({ version, expirationDate }) => {
        if (!semver.valid(version)) {
          logger.warn(`Skipped machine image ${machineImage.name} as version ${version} is not a valid semver version`)
          return false
        }
        return true
      })
      versions.sort((a, b) => {
        return semver.rcompare(a.version, b.version)
      })

      const name = machineImage.name
      const vendorName = vendorNameFromImageName(machineImage.name)
      const vendorHint = findVendorHint(configStore.vendorHints, vendorName)

      return map(versions, ({ version, expirationDate, cri, classification, architectures }) => {
        if (isEmpty(architectures)) {
          architectures = ['amd64'] // default if not maintained
        }
        return decorateClassificationObject({
          key: name + '/' + version,
          name,
          version,
          cri,
          classification,
          expirationDate,
          vendorName,
          icon: vendorName,
          vendorHint,
          architectures,
        })
      })
    })
  }
  
  function machineImagesByCloudProfileName (cloudProfileName) {
    const cloudProfile = cloudProfileByName(cloudProfileName)
    return get(cloudProfile, 'data.machineImages')
  }

[holgerkoser]

If both boolean flags are false we render an empty list item content block.

[grolu]

this cannot happen as the whole section is hidden in tis case see line 98

[petersutter]

I agree that it should not be logged (as an error) inside the getter. However, if there is no place where we log that we skipped some entries, this might make it harder to troubleshoot issues where an entry is expected to be there but is not.

[holgerkoser]

/lgtm

[petersutter]

/lgtm