1 If you can verify by sending only one machine, there is no need to publish all of them. Send one machine to verify immediately. If there is a problem, roll it back immediately, and continue to send if it passes.
2 The release of a function involves multiple systems. If the launch fails, it is not necessary for all systems to be rolled back.
For a specific system, if it is a new function and does not affect the old function, there is no need to roll back
Configuration files are divided into code-level configuration files, non-code-level configuration files,
If the code file is at the code level, it needs to be placed in the war package or jar package.
Non-code-level files need to be manually modified by operation and maintenance, such as machine ip, jdbc, redis address, specific IP
Operation and maintenance release process:
1 If the configuration file needs to be modified, the operation and maintenance shall uniformly modify (add, modify) the configuration file on the configuration file server (the operation and maintenance is called 'central control').
Configuration file subsystem, enterprise, goods, pms each have their own configuration file directory
2 Node machines build applications (war), etc., you need to take a unified configuration from the "configuration file server", sub-system, and customize the configuration as needed, for example, you need to configure your own IP
The customized configuration must be modified by script, the seed command cannot directly change the file, because the configuration file needs to be fetched from the central control every time, overwriting the original configuration
3 Start the node machines in turn to complete the release of the application