[GRM] Do not yield --help output on error

[ialidzhikov]

How to categorize this issue?
/area ops-productivity
/kind enhancement
/priority 3

What would you like to be added:

$ k -n garden logs gardener-resource-manager-57fd658c9-8lfkw -p

E0830 09:42:04.221217       1 leaderelection.go:325] error retrieving resource lock garden/gardener-resource-manager: context deadline exceeded
I0830 09:42:04.221260       1 leaderelection.go:278] failed to renew lease garden/gardener-resource-manager: timed out waiting for the condition
Error: error running manager: leader election lost
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.resource-controller","msg":"Shutdown signal received, waiting for all workers to finish"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.health-controller","msg":"Shutdown signal received, waiting for all workers to finish"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.secret-controller","msg":"Shutdown signal received, waiting for all workers to finish"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.garbage_collector","msg":"Shutdown signal received, waiting for all workers to finish"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.resource-controller","msg":"All workers finished"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.health-controller","msg":"All workers finished"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.garbage_collector","msg":"All workers finished"}
{"level":"info","ts":"2021-08-30T09:42:04.221Z","logger":"controller-runtime.manager.controller.secret-controller","msg":"All workers finished"}
Usage:
  gardener-resource-manager [flags]

Flags:
      --always-update                             if set to false then a resource will only be updated if its desired state differs from the actual state. otherwise, an update request will be always sent.
      --cache-resync-period duration              duration how often the controller's cache is resynced (default 24h0m0s)
      --cluster-id string                         optional cluster id for source cluster
      --garbage-collector-sync-period duration    duration how often the garbage collection should be performed (default: 0, i.e., gc is disabled)
      --health-bind-address string                bind address for the health server (default ":8081")
      --health-max-concurrent-workers int         number of worker threads for concurrent health reconciliation of resources (default 10)
      --health-sync-period duration               duration how often the health of existing resources should be synced (default 1m0s)
  -h, --help                                      help for gardener-resource-manager
      --kubeconfig string                         path to the kubeconfig for the source cluster
      --leader-election                           enable or disable leader election (default true)
      --leader-election-lease-duration duration   lease duration for leader election (default 15s)
      --leader-election-namespace string          namespace for leader election
      --leader-election-renew-deadline duration   renew deadline for leader election (default 10s)
      --leader-election-resource-lock string      Which resource type to use for leader election. Supported options are 'endpoints', 'configmaps', 'leases', 'endpointsleases' and 'configmapsleases'. (default "leases")
      --leader-election-retry-period duration     retry period for leader election (default 2s)
      --max-concurrent-workers int                number of worker threads for concurrent reconciliation of resources (default 10)
      --metrics-bind-address string               bind address for the metrics server (default ":8080")
      --namespace string                          namespace in which the ManagedResources should be observed (defaults to all namespaces)
      --resource-class string                     resource class used to filter resource resources (default "resources")
      --secret-max-concurrent-workers int         number of worker threads for concurrent secret reconciliation of resources (default 5)
      --sync-period duration                      duration how often existing resources should be synced (default 1m0s)
      --target-cache-resync-period duration       duration how often the controller's cache for the target cluster is resynced (default 24h0m0s)
      --target-disable-cache                      disable the cache for target cluster and always talk directly to the API server (defaults to false)
      --target-kubeconfig string                  path to the kubeconfig for the target cluster
      --version version[=true]                    Print version information and quit

[timebertt]

This should be sufficient: https://github.com/gardener/terraformer/blob/3cdb7d585d81a7b9cad9eb454a413cffb9a1897e/cmd/terraformer/app/terraformer.go#L100-L101

[acumino]

Similar change is required in -

gardenelet
gardener-scheduler
gardener-apiserver
gardener-controller-manager
gardener-resource-manager

Will all these be refactored like GSAC in #3459, then required changes can be made in that PR only and we can close this issue?
@rfranzke @timebertt

[rfranzke]

• gardener-scheduler and gardener-resource-manager are already native controller-runtime components.
• The plan is to also refactor gardener-controller-manager and gardenlet as part of Refactor Gardener components to native controller-runtime components #4251, however, this will probably take a bit more time. Hence, probably it makes sense to fix it already now if it is a simple change as proposed in [GRM] Do not yield --help output on error #4807 (comment).
• gardener-apiserver will not be refactored since it's no controller but an API server. There is https://github.com/kubernetes-sigs/apiserver-runtime but it's still experimental and there are no plans to adopt it.

[acumino]

• gardener-scheduler and gardener-resource-manager are already native controller-runtime components.
• The plan is to also refactor gardener-controller-manager and gardenlet as part of Refactor Gardener components towards native controller-runtime components #4251, however, this will probably take a bit more time. Hence, probably it makes sense to fix it already now if it is a simple change as proposed in [GRM] Do not yield --help output on error #4807 (comment).
• gardener-apiserver will not be refactored since it's no controller but an API server. There is https://github.com/kubernetes-sigs/apiserver-runtime but it's still experimental and there are no plans to adopt it.

Thanks, I will make the required changes.

[acumino]

/assign