-
Epic
-
Resolution: Unresolved
-
Normal
-
None
-
None
-
None
-
UI: ROSA HCP Cluster Autoscaler Support
-
False
-
-
False
-
To Do
-
XCMSTRAT-937 - ROSA HCP: Default Cluster Autoscaler Max Nodes
-
100% To Do, 0% In Progress, 0% Done
- User Story 1: Ability for cluster administrators to describe ROSA HCP cluster and view cluster autoscaler properties.
- Including both existing clusters and new clusters created after this feature is GA'ed.
- ROSA DESCRIBE AUTOSCALER command supports HCP clusters
- Default values for configurations are displayed
- A message is displayed when no Autoscaler is defined
- User Story 2: Ability for cluster administrators to define a CAS on an existing cluster when the cluster-autoscaler is disabled. (i.e., all the machine pools use manual scaling).
- Supported fields include
- MaxNodesTotal
- maxPodGracePeriod: With default of 600 seconds
- maxNodeProvisionTime: With default of 15 minutes
- podPriorityThreshold: default is -10
- Range for maxNodesTotal must be easily configurable/modifiable by the service as we increase scaling limits.
- User Story 3: Ability for a cluster administrator to delete a CAS on an existing cluster when the cluster-autoscaler is enabled.
- Allowed only when all the existing machine pools use manual scaling.
- User Story 4: Ability to own a cluster without cluster auto-scaler or define an auto-scaler that uses the defaults and configure auto-scaler with only valid values:
- maxNodesTotal:
- Default: must be same as that on existing clusters’ CAS
- Should be greater than 2 (current minimum no. of nodes required).
- Should not exceed the tested maximum (currently 500).
- Help message: MaxNodesTotal is the maximum allowable number of nodes across all MachinePools.
- maxPodGracePeriod:
- Default: 600 seconds or 10 minutes
- Minimum value supported 600 seconds
- Maximum value supported 0 (indefinite)
- maxNodeProvisionTime:
- Default: 15 minutes
- Permissible values: [15m - 60m]
- podPriorityThreshold:
- Default : -10
- Permissible values: [-2147483648 to 1000000000]
- User Story 5: Ability to create a new machine pool or edit an existing machine pool with Cluster Autoscaler configured.
- Issue an informational message that ‘Scaling out to maximum replicas is subjected to cluster autoscaler configuration’.
- When the sum of replicas (if autoscaling is disabled) and max-replicas (if autoscaling is enabled) exceeds the MaxNodesTotal, issue a warning message that ‘Actual maximum replicas can be lower since the maximum replicas defined exceeds Cluster Autoscaler limit (MaxNodes).’
- When the sum of replicas (if autoscaling is disabled) and min-replicas (if autoscaling is enabled) exceeds the MaxNodesTotal, issue a warning message that ‘Actual total nodes in the cluster will be more than the maximum nodes configured in the cluster autoscaler’.
- When the sum of replicas (if autoscaling is disabled) and min-replicas (if autoscaling is enabled) exceeds the tested maximum (i.e. 90 or 500), issue an error message that ‘The total number of nodes in the cluster exceeds the supported maximum’.
- Issue an error message that ‘Minimum replicas configured exceeds the maximum nodes defined in Cluster Autoscaler’ when min-replicas makes total cluster nodes > maxNodes.