Understanding Memory Usage in PlanetScale: Cache vs. Process Memory

The article explains that the memory usage figure shown in the PlanetScale dashboard is a combination of active cache, inactive cache, RSS, and memory‑mapped pages, not just Postgres data. Active cache is reclaimable by the OS and indicates that frequently accessed data is kept close to the CPU, which is healthy for performance. RSS, or Resident Set Size, represents non‑reclaimable memory used by Postgres processes and is the main indicator of out‑of‑memory risk. The author clarifies that high memory usage can be healthy if cache dominates, but high RSS signals memory pressure. Postgres uses two caching layers: its own shared_buffers pool and the OS page cache, both of which consume RAM. The work_mem parameter, defaulted relative to cluster memory, can be tuned to control per‑query memory allocation, but setting it too low forces disk spills while setting it too high increases RSS. The article recommends monitoring metrics via the PlanetScale dashboard, checking the 'Normal operating ranges' documentation, and ensuring that cache usage stays high while RSS remains low. Understanding the distinction between cache and process memory helps prevent unnecessary restarts and performance degradation.