I have deployed an Event Hub triggered Azure Function written in Java on AKS. The function should scale out using KEDA. The function is correctly triggerd and working but it's not scaling out when the load increases. I have added sleep calls to the function implementation to make sure it's not burning through the events too fast and should be forced to scale out but this did not show any change as well.
kubectl get hpa shows the following output
NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS AGE
keda-hpa-eventlogger Deployment/eventlogger 64/64 (avg) 1 20 1 3m41s
This seems to be a first indicator that something is not correct as i assume the first number in the targets column is the number of unprocessed events in event hub. This stays the same no matter how many events i pump into the hub.
The Function was deployed using the following Kubernetes Deployment Manifest
data:
AzureWebJobsStorage: <removed>
FUNCTIONS_WORKER_RUNTIME: amF2YQ==
EventHubConnectionString: <removed>
apiVersion: v1
kind: Secret
metadata:
name: eventlogger
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: eventlogger
labels:
app: eventlogger
spec:
selector:
matchLabels:
app: eventlogger
template:
metadata:
labels:
app: eventlogger
spec:
containers:
- name: eventlogger
image: <removed>
env:
- name: AzureFunctionsJobHost__functions__0
value: eventloggerHandler
envFrom:
- secretRef:
name: eventlogger
readinessProbe:
failureThreshold: 3
periodSeconds: 10
successThreshold: 1
timeoutSeconds: 240
httpGet:
path: /
port: 80
scheme: HTTP
startupProbe:
failureThreshold: 3
periodSeconds: 10
successThreshold: 1
timeoutSeconds: 240
httpGet:
path: /
port: 80
scheme: HTTP
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: eventlogger
labels:
app: eventlogger
spec:
scaleTargetRef:
name: eventlogger
pollingInterval: 5
cooldownPeriod: 5
minReplicaCount: 0
maxReplicaCount: 20
triggers:
- type: azure-eventhub
metadata:
storageConnectionFromEnv: AzureWebJobsStorage
connectionFromEnv: EventHubConnectionString
---
The Connection String of Event Hub contains the "EntityPath=" Section as described in the KEDA Event Hub Scaler Documentation and has Manage-Permissions on the Event Hub Namespace.
The output of kubectl describe ScaledObject is
Name: eventlogger
Namespace: default
Labels: app=eventlogger
scaledobject.keda.sh/name=eventlogger
Annotations: <none>
API Version: keda.sh/v1alpha1
Kind: ScaledObject
Metadata:
Creation Timestamp: 2022-04-17T10:30:36Z
Finalizers:
finalizer.keda.sh
Generation: 1
Managed Fields:
API Version: keda.sh/v1alpha1
Fields Type: FieldsV1
fieldsV1:
f:metadata:
f:annotations:
.:
f:kubectl.kubernetes.io/last-applied-configuration:
f:labels:
.:
f:app:
f:spec:
.:
f:cooldownPeriod:
f:maxReplicaCount:
f:minReplicaCount:
f:pollingInterval:
f:scaleTargetRef:
.:
f:name:
f:triggers:
Manager: kubectl-client-side-apply
Operation: Update
Time: 2022-04-17T10:30:36Z
API Version: keda.sh/v1alpha1
Fields Type: FieldsV1
fieldsV1:
f:metadata:
f:finalizers:
.:
v:"finalizer.keda.sh":
f:labels:
f:scaledobject.keda.sh/name:
f:status:
.:
f:conditions:
f:externalMetricNames:
f:lastActiveTime:
f:originalReplicaCount:
f:scaleTargetGVKR:
.:
f:group:
f:kind:
f:resource:
f:version:
f:scaleTargetKind:
Manager: keda
Operation: Update
Time: 2022-04-17T10:30:37Z
Resource Version: 1775052
UID: 3b6a68c1-c3b9-4cdf-b5d5-41a9721ac661
Spec:
Cooldown Period: 5
Max Replica Count: 20
Min Replica Count: 0
Polling Interval: 5
Scale Target Ref:
Name: eventlogger
Triggers:
Metadata:
Connection From Env: EventHubConnectionString
Storage Connection From Env: AzureWebJobsStorage
Type: azure-eventhub
Status:
Conditions:
Message: ScaledObject is defined correctly and is ready for scaling
Reason: ScaledObjectReady
Status: False
Type: Ready
Message: Scaling is performed because triggers are active
Reason: ScalerActive
Status: True
Type: Active
Status: Unknown
Type: Fallback
External Metric Names:
s0-azure-eventhub-$Default
Last Active Time: 2022-04-17T10:30:47Z
Original Replica Count: 1
Scale Target GVKR:
Group: apps
Kind: Deployment
Resource: deployments
Version: v1
Scale Target Kind: apps/v1.Deployment
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal KEDAScalersStarted 10s keda-operator Started scalers watch
Normal ScaledObjectReady 10s keda-operator ScaledObject is ready for scaling
So i'm a bit stucked as i don't see any errors but it's still not behaving as expected.
Versions:
Was able to find a solution to the problem.
Event Hub triggered Azure Functions deployed on AKS show the same scaling characteristics as Azure Functions on App Service show:
You only get one consumer per partition to allow for ordering per partition. This characteristic overrules maxReplicaCount in the Kubernetes Deployment Manifest.
So to solve my own issue: By increasing the partitions for Event Hub i get a pod per partition and KEDA scales the workload as expected.