Describe the enhancement:
currently when using autodiscover for pod resources, beats adds deployment.name if the specific pod owned by ReplicaSet that's also owned by Deployment, in this function getRSDeployment, go client was used directly without leveraging k8s cache/store, however go client has QPS 5 and Burst 10 by default, which could be a bottle neck of autodiscover. It may not be a good idea to introduce a replicaset watcher because of memory overhead, which seems the only way to accelerate this function is adding options to configure the client's QPS/Burst.
Describe a specific use case for the enhancement or feature:
in k8s clusters that a single beats' pod is responsible for a lot of pods.
Describe the enhancement:
currently when using autodiscover for
podresources, beats addsdeployment.nameif the specific pod owned byReplicaSetthat's also owned byDeployment, in this function getRSDeployment, go client was used directly without leveraging k8s cache/store, however go client has QPS 5 and Burst 10 by default, which could be a bottle neck of autodiscover. It may not be a good idea to introduce a replicaset watcher because of memory overhead, which seems the only way to accelerate this function is adding options to configure the client's QPS/Burst.Describe a specific use case for the enhancement or feature:
in k8s clusters that a single beats' pod is responsible for a lot of pods.