Kubernetes version: 1.14.2
Cloud being used: bare-metal
CNI and version: weave 2.5.2
CRI and version: Docker 18.09.1
I have notices some strange network performance problems with my cluster.
It seems that our K8s Cluster, with 1 Master and 4 Nodes is slower as Minikube on my local machine.
When I do a Test on an application like Wirecloud I have massive network time delays.
As example: http://my.server.name:32000/myapp
For a better understanding I separated the problem description in 2 Phases.
- It takes about 4-5 seconds until the application begins to load
(I think this is called time to first byte)
- Application loads 2-3 seconds
I see major differences in Phase 1. As example in our production and test environment I have timeslots from 4-5 seconds. When I install Minikube the time to the first byte is only 0-1 seconds. All the server are installed with Cent Os 7.x.latest.
We are using the weave-net pod version 2.5.2 and Kubernetes Version 1.13.4 (Prod) and 1.14.2) on the PreProd. The System is installed in an VM environment managed via VMWare.
I did some research about different tuning profiles, but in our situation this is set by default
- virtual-guest - Optimize for running inside a virtual guest
So for my understanding the difference is only in the network, has anyone a quick shot or a lucky finding for me ?