Loading...

XML

Word

Printable

Type: Bug
Resolution: Done
Priority: Undefined
Fix Version/s: None
Affects Version/s: None
Labels:
None

Blocked:
False
Blocked Reason:
None
Ready:
False
Intelligence Requested:
Market:

Severity:
Moderate

SFDC Cases Links:
SFDC Cases Counter:
SFDC Cases Open:

Summary

Using maxReplicas less than the number of MachineSets can cause problems with MachinePools.

Steps to Reproduce

Deploy a cluster to AWS. For the MachinePool, specify 3 AZs similar to the following:

apiVersion: hive.openshift.io/v1
kind: MachinePool
metadata:
  name: kevin-3-zone-worker
  namespace: kevin-3-zone
spec:
  clusterDeploymentRef:
    name: kevin-3-zone
  name: worker
  platform:
    aws:
      rootVolume:
        iops: 2000
        size: 100
        type: io1
      type: m5.xlarge
      zones:
      - us-east-1a
      - us-east-1b
      - us-east-1c
  replicas: 3

After the cluster has deployed, enable auto-scaling on the MachinePool with maxReplicas less than the number of MachineSets by adding the following spec and removing the replicas count:

spec:
  autoscaling:
    maxReplicas: 2
    minReplicas: 1

Expected Results

At least 1 node is removed from the cluster, and the MachinePool status is updated to show the new number of replicas.

Actual Results

A node is removed from the cluster, but the MachinePool status is not updated:

status:
...
  machineSets:
  - maxReplicas: 1
    minReplicas: 1
    name: kevin-3-zone-m5pfj-worker-us-east-1a
    readyReplicas: 1
    replicas: 1
  - maxReplicas: 1
    minReplicas: 1
    name: kevin-3-zone-m5pfj-worker-us-east-1b
    readyReplicas: 1
    replicas: 1
  - maxReplicas: 1
    minReplicas: 1
    name: kevin-3-zone-m5pfj-worker-us-east-1c
    readyReplicas: 1
    replicas: 1
  replicas: 3

Additional Information

There may be a failure when Hive attempts to create a MachineAutoscaler on the cluster with maxReplicas set to 0 that prevents the status of the MachinePool from being updated. See Slack discussions in #forum-hive and #forum-ocm-cloud