Foreword
Relatively CentOS, I'm more familiar with Ubuntu, but it comes with Python3.6, there is a very detailed ssh login prompt after completion so I deployed a large number of virtual machines in Ubuntu development environment.
But did not last long, Jenkins Slave and Kubernetes Worker have problems. .
Problem Description
The first problem is that our main Jenkins Slave, in Ubuntu virtual machine inside, using Docker ran Jenkins Slave four different environments, provide c # / golang / flutter / python like build / test environment.
Over time, this Jenkins Slave virtual machine's memory is high. This is probably some building maintenance tasks daemon to speed up construction of the next, so the memory is not freed.
Then the question becomes, obviously prompt the virtual machine as well as 2G of free memory (does not contain cached), but ssh login can spend up to two or three minutes after login top / htop also black almost a minute. .
Amazingly other commands (ls / cd / df / du / docker) are nothing but normal, no obvious Caton phenomenon.
Another Kubernetes Worker is the same problem, but also more than half the memory, CPU usage does not exceed 50%, can ssh / top / htop these commands cards may not work. Running on the node Nginx-Ingress-Controller directly stuck, you must restart the node to function properly. .
Cause is not found