LinkedIn
Twitter
I have a cloud run endpoint that gets triggered by a push type pubsub subscription and the cloudrun api is a composite api which does 2-3 operations and then it returns. I am observing at high load times few of the third party things like redis and firestore produce timeouts. In that case I am expecting pubsub will control the load (as per the document -> delivery rate) since there will be nacks coming from service. All the requests are getting successful in push backoffs but my service is experiencing the failures. Please help me what i am missing or how to monitor the push window?
I tried to reproduce the scenario where i was pushing 1000 messages to pubsub and my cloudrun instance was set to max 1 with 1 concurrency and 10secs timeout. I observed the send_request_count is increasing first then it settles but as per the slow start algorithm it should increase gradually based on the load handling capacity. Attached the graph below.
Blue line -> send_request_count
Red line ->. No of unreachable request
Red line ->. No of unreachable request
