timeout_seconds seems to be ineffective

[wei1793786487]

I set timeout_seconds to 600 seconds, but he still completes the watch event every 60 seconds. Can anyone provide assistance

 async def watch_pods(self):
            logger.info("Watching pods...")
            await self.config_loaded
            while True:  
                try:
                    async with client.ApiClient() as api:
                        v1 = client.CoreV1Api(api)
                        w = watch.Watch()
                        async for event in w.stream(v1.list_pod_for_all_namespaces, watch=True, timeout_seconds=600,):
                            pod = event['object']
                            pod_name = pod.metadata.name
                            self.resource_version = pod.metadata.resource_version
                            pod_namespace = pod.metadata.namespace
                            event_type = event['type']
                            if pod_namespace not in CHECK_NAMESPACES:
                                continue
                            if event_type in ['ADDED', 'MODIFIED']:
                                if pod.status.phase in ['Running', 'Pending']:
                                    self.pod_cache[f'{pod_namespace}/{pod_name}'] = pod
                                    logger.info(f"Pod {pod_name} in namespace {pod_namespace} {event_type.lower()}.")
                            elif event_type == 'DELETED':
                                self.pod_cache.pop(f'{pod_namespace}/{pod_name}', None)
                                logger.info(f"Pod {pod_name} in namespace {pod_namespace} deleted.")
                            logger.info("Pod watch stream ended.")
                except Exception as e:
                    logger.exception(e)
                    logger.error(f"Error watching pods: {e}")
                    logger.info("Reconnecting after error...")
                    await asyncio.sleep(1)

[tomplus]

Thank for reporting the issue. Could you check if you example is correct - it looks like this line "Pod watch stream ended." is printed when you get delete events...

Which version of K8s do you use? Is it self-hosted or managed by a cloud provider like Google, AWS, Alibaba Cloud,... ?

[wei1793786487]

Sorry, I had an indentation issue, but I've fixed my code. I am using a self-hosted Kubernetes version v1.21.5, and I am new to Kubernetes. I want to know if the problem is with my code logic or with my Kubernetes setup

[tomplus]

Could you check your code with the latest version of the library (32.0.0)? There are some fixes related to timeouts. Thank you.