Summary of TCP Troubleshooting Cases
I have encountered all kinds of strange TCP-related problems, so I will summarize and record them. By analyzing all the ins and outs of these problems, you can help you establish several solid grippers in the TCP knowledge system, so that TCP knowledge can slowly grow and communicate between grippers
Cause analysis of the phenomenon that the service does not respond or the strange abnormality
Analysis of the strange behavior of a black box program The full connection queue on the listen port soon overflowed, causing the entire program to not respond
Take three inferences – from theoretical knowledge to practical problems, a large number of CLOSE_WAIT appear on the server side, which is exactly equal to somaxconn (CLOSE_WAIT will also become the same value after adjusting the size of somaxconn)
See you for a long time, the TCP connection is connected to each other and the application always throws several connection exception errors after a period of time, and the cause needs to be found out. After investigation, it was found that the TCP connections were connected to each other. This case is really cherished, so I recorded it.
Transmission speed analysis
Case: TCP transmission speed case analysis (long fat network, rt increase, the impact of delay ack, etc.)
Principle: You need to understand the relationship between TCP-performance and sending and receiving Buffer: the impact of sending window size (Buffer) and receiving window size (Buffer) on TCP transmission speed, and how to observe the impact of window on transmission speed. How do BDP, RT, and bandwidth affect transmission speed?
TCP queue issues and number of connections
Firewall and reset location analysis
Use of ttl, identification, etc.
TCP related parameters
Tools and Skills
This article is reprinted from https://blog.p2hp.com/archives/9444
This site is for inclusion only, and the copyright belongs to the original author.