Kubernetes abnormal communication network fault solution ideas

stay k8s The most important thing is the network , If the network is not available , Then the cluster is equivalent to the scrap state ,

 

 

Cause of network failure  

---

Kubernetes The network in is very important , Whole k8s The communication of the cluster depends on the network plug-in , Such as calico、flannel、canel etc. , Communication is achieved through network plug-ins ,kubernetes The most common faults in the network communication process above are as follows ∶（calico It can be for pod Distribute ip Address ,calico be relative to flannel You can also configure network policies ）

Communication mainly includes the following ∶

• Pod Own network failure
• Pod and Pod Communication failure （ On the same node pod Communication and cross node pod signal communication ）
  Pod Communication failure with physical node （pod Unable to access physical nodes IP Or physical nodes ping pod Of ip yes ping It doesn't work ）

• Communication between containers

The same pod Communication between multiple containers inside ,lo, Through the local loopback network card lo communicate

• pod signal communication

From a pod lp To another pod Ip

• Pod And service External communication

pod ip To cluster ip

• External communication

Service Communication with clients outside the cluster , If you use service Realization 4 Layer load balancing , Want to access outside the cluster service The associated pod, Need to put service Turn into nodeport type , Only in this way can the host be used outside the cluster IP Access with the exposed port of the host service Internal application .