[NEW] Scan with a specified probability

[guozhihao-224]

The problem/use-case that the feature addresses

In scenarios like data migration validation or hotkey analysis, users need to efficiently sample a subset of keys without full-database scanning. Current approaches require either:

• Pulling all keys via SCAN and filtering client-side (high network overhead)， It has a large impact on the original node, and the full scan takes a long time
• Running full RDB checks (resource-intensive for large datasets)， the operation needs to be performed on the machine of the instance

Description of the feature

we can enhan the functionality of the scan command：

SCAN cursor [MATCH pattern] [PROBABILITY float] [COUNT count]

• PROBABILITY: 0.0-1.0 (probability threshold)

[zuiderkwast]

Interesting idea. :) I have also seen request to extend RANDOMKEY with a count argument.

You could call SCAN from a Lua script and discard elements based on probability in Lua code.

[guozhihao-224]

Interesting idea. :) I have also seen request to extend RANDOMKEY with a count argument.

You could call SCAN from a Lua script and discard elements based on probability in Lua code.

Yes, but if do this, each instance may have to upload the lua script when it starts, and an additional place is needed to save the lua script. This might not be so convenient.