Greptile Overview

Greptile Summary

This PR fixes metrics parquet file searching in WAL by properly handling partition time levels (daily vs hourly), and refactors cache control from negative no_cache to positive use_cache semantics.

Key Changes:

• Fixed WAL parquet search to respect partition_time_level setting, correctly aligning time ranges to either daily or hourly boundaries based on stream configuration
• Refactored cache control from no_cache to use_cache throughout the codebase for clearer intent
• Added cache update logic that allows queries with caching disabled to still update existing cache entries
• Implemented frontend support for window.use_cache override to control query caching behavior during development/debugging

Issue Found:

• Comment on line 100 in src/service/promql/search/mod.rs incorrectly describes the logic (says "use_cache is true" when it should say "use_cache is false")

Confidence Score: 4/5

• This PR is safe to merge with one minor documentation fix needed
• The changes are well-structured and address the stated issues. The partition time level fix is correct and properly threaded through the codebase. The cache semantics refactor is consistent across all files. Only issue is an incorrect comment that doesn't affect functionality
• src/service/promql/search/mod.rs needs comment correction on line 100

Important Files Changed

File Analysis

Sequence Diagram

sequenceDiagram
    participant Client as Web Client
    participant HTTP as HTTP Handler
    participant Search as PromQL Search
    participant Cache as Cache Service
    participant Cluster as Cluster RPC
    participant WAL as WAL Search
    participant Parquet as Parquet Files

    Client->>HTTP: query_range(query, use_cache)
    HTTP->>Search: search_in_cluster(req)
    
    Note over Search: Check if cache_disabled<br/>(!use_cache || !cache_enabled)
    
    alt Cache enabled and use_cache=true
        Search->>Cache: get(query, start, end)
        Cache-->>Search: cached data (if exists)
    end
    
    Search->>Cluster: distribute query to nodes
    Cluster->>WAL: search_parquet(query)
    
    Note over WAL: Get partition_time_level<br/>from stream settings
    
    alt Daily Partitioning
        WAL->>WAL: Align to DAY_MICRO_SECS
    else Hourly Partitioning
        WAL->>WAL: Align to HOUR_MICRO_SECS
    end
    
    WAL->>Parquet: scan files in time range
    Parquet-->>WAL: record batches
    WAL-->>Cluster: results
    Cluster-->>Search: aggregated results
    
    alt Cache enabled
        Search->>Cache: set(query, results, update=!cache_disabled)
        Note over Cache: Update cache if needed
    end
    
    Search-->>HTTP: query results
    HTTP-->>Client: response

_{12 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}