In highly concurrent projects, when is the cache, opcode block, server, and database configuration?

The project is a concurrency of tens of thousands, and the data volume reaches tens of millions. the architecture is LNMP, with a redis and opcode. For example, cache (redis), database configuration (MySQL), and server configuration (linux system configuration) are added to the project during the development process, wait until the project is complete... the project is a concurrency of tens of thousands, and the data volume reaches tens of millions. the architecture is LNMP, with a redis and opcode.

For example, cache (redis), database configuration (MySQL), and server configuration (linux system configuration) are added to the project during the development process, or after the project is completed, can it be reconstructed? How can we weigh them?

My concerns are:
If I add a cache (redis) to the project, I will not encounter many problems during the debugging process? For example, if the data synchronization is inconsistent and the code is modified, no response is returned.
However, if the reconstruction is completed after the project is completed, the cost may seem a little high?

If you have experience, give me some advice. Thank you.

Reply content:

The project is a concurrency of tens of thousands, and the data volume reaches tens of millions. the architecture is LNMP, with a redis and opcode.

For example, cache (redis), database configuration (MySQL), and server configuration (linux system configuration) are added to the project during the development process, or after the project is completed, can it be reconstructed? How can we weigh them?

My concerns are:
If I add a cache (redis) to the project, I will not encounter many problems during the debugging process? For example, if the data synchronization is inconsistent and the code is modified, no response is returned.
However, if the reconstruction is completed after the project is completed, the cost may seem a little high?

If you have experience, give me some advice. Thank you.

My suggestion is: make sure that the production environment and development environment are consistent, so that some problems can only occur online.
For configuration, you can configure an intranet and an Internet.
You can deploy redis and mysql on the intranet, and then write an intranet configuration. after the configuration is online, use the Internet configuration.
If you do not do this, it is very dangerous to change the configuration online.