关于Redis模块可定制范围及底层内存分配函数替换可行性的咨询

Hey there! Let's walk through your question about customizing Redis's low-level memory handling via its module system.

So in an ideal world , I would like to replace some very low level stuff . For instance the 'zmalloc' , some of them I want to replace with my own version of malloc , and I also get to know which areas of memory has been changed and so on .There is little chance thats possible only via Modules , but if it was , it would be great .

First off, you’re spot-on that replacing core low-level functions like zmalloc using only Redis Modules is extremely difficult—if not impossible for practical, stable use. The Redis Module API is built to extend high-level functionality: think custom commands, new data types, or hooking into specific runtime events. It doesn’t expose hooks to swap out the foundational memory allocation primitives that the entire Redis server relies on. These allocators are deeply integrated into Redis’s core codebase, and there’s no supported way for modules to override them without risking crashes, memory leaks, or undefined behavior.

I appreciate you experimenting with undefining zmalloc and rolling your own version, plus using gdb to trace Redis’s internal calls—that’s a great way to dig into how the internals work! But even if you get something to work in a test environment, it’s not a sustainable approach for production. Redis’s internals don’t account for third-party code hijacking these core allocators, so you’d run into all sorts of edge cases: misaligned memory, mismatched allocation/free patterns, or conflicts with Redis’s own built-in memory tracking.

If tracking memory changes is a key priority, you might want to start with Redis’s built-in tools first:

• Use the INFO memory command to get a high-level overview of memory usage
• Run MEMORY USAGE <key> to check the memory footprint of individual keys
• Enable Redis’s memory event notifications (if applicable) to get alerts about key evictions or memory limits

If you really need custom memory tracking or a custom allocator, your best bet would be to patch Redis’s core code directly. That lets you integrate your own zmalloc replacement and tracking logic directly into the server, but keep in mind this means maintaining your own fork of Redis, which adds ongoing maintenance work to keep up with upstream updates.

备注：内容来源于stack exchange，提问作者Mohamed Mchaouri