I am getting this error when deploy kafka => Ignoring pod, pod anti-affinity is not supported. But some kafka pod deployed.

[burakhalefoglu]

my provisioner;
apiVersion: karpenter.sh/v1alpha5
kind: Provisioner
metadata:
name: default
spec:
limits:
resources:
cpu: "1000"
memory: 1000Gi
requirements:
- key: "node.kubernetes.io/instance-type"
operator: In
values: ["r5n.large","r5n.xlarge","r5n.2xlarge","r5n.4xlarge", "t3.large","t3.xlarge","t3.2xlarge","t3.4xlarge", "m5n.large","m5n.xlarge","m5n.2xlarge","m5n.4xlarge","c5n.xlarge","c5n.2xlarge","c5n.4xlarge"]
- key: "topology.kubernetes.io/zone"
operator: In
values: ["us-east-2a", "us-east-2b", "us-east-2c"]
- key: "karpenter.sh/capacity-type" # Defaults to on-demand
operator: In
values: ["spot"] # ["spot", "on-demand"]
provider:
subnetSelector:
karpenter.sh/discovery: xxx
securityGroupSelector:
karpenter.sh/discovery: xxx
ttlSecondsAfterEmpty: 30
ttlSecondsUntilExpired: 2592000

---

my cluster config ->

apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig
metadata:
name: xxx
region: us-east-2
version: '1.21'
tags:
karpenter.sh/discovery: xxx
managedNodeGroups:

• name: xxx
  desiredCapacity: 2
  labels:
  dedicated: spot
  volumeSize: 20
  amiFamily: AmazonLinux2
  minSize: 1
  maxSize: 20
  spot: true
  privateNetworking: true
  volumeEncrypted: true
  availabilityZones: ["us-east-2a","us-east-2b","us-east-2c"]
  instanceTypes: ["c5n.xlarge","c5n.2xlarge","c5n.4xlarge","r5n.xlarge","r5n.2xlarge","r5n.4xlarge","m5n.xlarge","m5n.2xlarge","m5n.4xlarge"]
  iam:
  withAddonPolicies:
  autoScaler: true
  ssh:
  allow: false

---

kafka:
afka kafka-zookeeper-0 0/1 Pending 0 0s
kafka kafka-0 0/1 Pending 0 0s
kafka kafka-1 0/1 Pending 0 0s
kafka kafka-2 0/1 Pending 0 0s
kafka kafka-2 0/1 Pending 0 0s
kafka kafka-3 0/1 Pending 0 0s
kafka kafka-3 0/1 Pending 0 0s
kafka kafka-4 0/1 Pending 0 0s
kafka kafka-4 0/1 Pending 0 0s
kafka kafka-zookeeper-0 0/1 Pending 0 6s
kafka kafka-0 0/1 Pending 0 6s
kafka kafka-zookeeper-0 0/1 ContainerCreating 0 6s
kafka kafka-0 0/1 ContainerCreating 0 6s
kafka kafka-1 0/1 Pending 0 6s
kafka kafka-1 0/1 ContainerCreating 0 6s
logging elasticsearch-master-2 0/1 Pending 0 20m
kafka kafka-1 0/1 Running 0 26s
kafka kafka-1 0/1 Error 0 36s
kafka kafka-1 0/1 Running 1 37s
kafka kafka-zookeeper-0 0/1 Running 0 37s
kafka kafka-0 0/1 Running 0 38s
kafka kafka-zookeeper-0 1/1 Running 0 46s
kafka kafka-1 0/1 Error 1 47s
kafka kafka-0 0/1 Error 0 47s
kafka kafka-0 0/1 Running 1 48s
kafka kafka-0 1/1 Running 1 56s
kafka kafka-1 0/1 CrashLoopBackOff 1 56s
kafka kafka-1 0/1 Running 2 57s
kafka kafka-1 1/1 Running 2 66s

kafka-0, kafka-1 is deployed but. kafka-2,-3,-4 stay pending.

I have 2 node now.
ip-xx-xx-xx-xx.us-east-2.compute.internal Ready 36m
ip-xx-xx-xx-xx.us-east-2.compute.internal Ready 36m

Version

Karpenter: v0.6.5

Kubernetes: v1.21.x

Expected Behavior

Actual Behavior

Steps to Reproduce the Problem

Resource Specs and Logs

[dewjam]

Hey @burakhalefoglu ,
One thing to note about Karpenter is it will only add new nodes when pods are unable to be scheduled (in Pending state). So if there is sufficient capacity in the cluster which meet a pods requirements, kube-scheduler will use those existing nodes first before Karpenter can take any action.

In your case, I would imagine that kafka-0 and kafka-1 are being scheduled on the existing managed nodes. But Karpenter is ignoring kafka-2 through kafka-5 because Pod ant-affinity is not currently supported (#942).

Can you run this command to confirm?

kubectl get pods -o wide -n <kafka_namespace>

Also, this may help us understand why pods are still pending.

kubectl describe pod <kafka_pod> -n <kafka_namespace>

[dewjam]

Hey @burakhalefoglu . Any updates on this issue? Anything else we can help with?

[dewjam]

Hey @burakhalefoglu . I haven't heard back on this issue for a while now. I'm going to go ahead and close it. Feel free to reopen if you still have questions.

Thanks!