Container Apps Job: KEDA scale rule fires continuously but no executions are created (no ContainerAppController entries)

Question

Container Apps Job: KEDA scale rule fires continuously but no executions are created (no ContainerAppController entries)

Aria Rahujani 20

My Container Apps Job (broker-discovery-job, East US 2, Workload profiles / Consumption, KEDA 2.18.1) has an azure-queue scale rule pointing at an Azure Storage queue.

KEDA is firing correctly — every 5 minutes it logs:

MetricValue: 1.00, CalculatedJobsNeeded: 1, Active: true

in ContainerAppSystemLogs_CL, for over 90 minutes straight.

But no execution is ever created. Execution history is empty. Zero ContainerAppController entries appear in the system log for the same window — no pod scheduled, no image pulled, nothing.

Already verified:

Queue has a visible message (Dequeue count 0)
storage-conn secret has the correct connection string (refreshed from key1)
Scale rule points to the right account and queue
Workload profile has capacity (0 instances in use, max 4 vCPUs / 8 GiB)
Environment status: Succeeded, no warnings
Suspend/Resume cycled, no effect
Queue deleted and recreated to clear any phantom invisible messages

Where should I look for the error that must be happening when the controller tries to create the execution? Nothing surfaces in the job's logs or activity log.

0 comments

1 answer

Your answer

Answer 1

Hi Aria Rahujani & thx for join me at Q&A platform,

seems like a problem between KEDA detecting the queue event and the Container Apps Jobs controller actually creating an execution. The important part is MetricValue: 1, CalculatedJobsNeeded: 1, and Active: true. KEDA is clearly seeing the message and calculating that a job is needed. Also, Dequeue count: 0 is expected here, bc checking queue length doesn't consume the message.

I'd check the job's actual triggerType and eventTriggerConfig.scale, especially maxExecutions, using az containerapp job show. Then try az containerapp job start manually. That's probably the most useful isolation test rn. If a manual execution starts normally but KEDA-triggered executions never appear, the issue is likely in the event-driven job scheduling/reconciliation path rather than the container image, queue credentials, or workload capacity. The missing ContainerAppController entries are interesting too. They don't necessarily prove the controller never received the request, but they suggest the failure may happen before an execution resource is created. Unfortunately, those internal controller decisions aren't fully exposed through customer-accessible logs.

If manual execution works and the scaling configuration is valid, I'd escalate this to the Azure Container Apps Jobs engineering team. Give them the job/environment resource IDs, region, KEDA logs with UTC timestamps, and the empty execution history. Ask them to inspect the internal ScaledJob-to-execution reconciliation. At that point, recreating the queue or refreshing secrets again probably won't help. U need Microsoft to check why KEDA's calculated demand isn't turning into actual job executions

rgds,

Alex