QuestionQ70

VerifiedAnalyzing and optimizing technical and business processes

Your company runs a Kubernetes application that retrieves messages from Pub/Sub and writes them to Filestore. Because the application is simple, it was deployed as one pod. The infrastructure team analyzed Pub/Sub metrics and found that the application cannot process messages in real time; most wait minutes before processing. You need to scale the I/O-intensive elaboration process. What should you do?

Explanation

A Horizontal Pod Autoscaler should scale the consumer deployment from the Pub/Sub subscription backlog. subscription/num_undelivered_messages reports the number of unacknowledged messages waiting in the subscription, directly indicating that consumers are not keeping up. It is more suitable than CPU utilization for an I/O-intensive workload, because CPU usage may not reflect processing saturation.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!