VersionAppBloomService

Info

Redis-backed bloom filter service for fast app name existence checks, preventing cache penetration on version lookups.

Overview

VersionAppBloomService implements a two-layer probabilistic filter: a Redis bitmap bloom filter for fast rejection, and a Redis Set for exact membership verification. This eliminates false positives from the bloom filter alone. Redis failures use fail-open semantics (return true) to avoid blocking legitimate lookups.

Methods

Method	Description
mightContain(appName)	Returns true if appName might exist (bloom + exact set check); false if definitely absent
add(appName)	Adds appName to both the bloom bitmap and the exact set
remove(appName)	Removes appName from the exact set only (bitmap bits are retained)
rebuildFromAppNames(appNames)	Clears and rebuilds both structures from a collection of app names

Two-Layer Check

mightContain("my-app")
  ├─ Bloom bitmap: any bit == 0 → return false (definitely absent)
  ├─ Bloom bitmap: all bits == 1 → check exact set
  │   ├─ Exact set: member → return true
  │   └─ Exact set: not member → return false (bloom false positive)
  └─ Redis error → return true (fail-open)

Configuration

Property	Default	Description
version.bloom.key	version:app:bloom	Redis key for the bloom bitmap
version.app.set.key	version:app:set	Redis key for the exact set
version.bloom.bit-size	1000003	Bitmap size (min 1024)
version.bloom.hash-count	6	Number of hash functions (min 2)

Hash Strategy

Uses double hashing with CRC32 and Java hashCode:

offset_i = (hash_crc32 + i * hash_java + i^2) % bitSize

Tips

The exact Set acts as a secondary filter to eliminate bloom false positives, ensuring mightContain returns false only when the appName is truly absent.