Parsing GC Logs can be slow

[karianna]

For some logs, parsing is slow. There are a myriad of factors involved in this, including the order or rules and how they are applied. The ZGC log attached with permission from AliBaba is an example of this.

slow_zgc_log.log

[karianna]

Kirk has a G1 log that he is baselining

[dsgrieve]

PR #295 gives about a 28% improvement, but I wouldn't consider this problem solved.

[brunoborges]

/relevance-check

[github-actions]

Relevance Assessment: Still Relevant

• Summary: This issue tracks a general parsing performance problem that has been partially addressed through multiple PRs but remains an open concern, with active work still ongoing.

• Evidence:

  • PR Order keys by mru parse rule #295 (merged June 2023) introduced MRU (Most Recently Used) rule ordering in RuleSet.java, yielding a ~28% improvement on a 394MB G1GC log. The contributor @dsgrieve explicitly stated: "I wouldn't consider this problem solved."
  • PR Optimize decorator parsing #458 (merged Nov 2025) optimized decorator parsing by folding the tags regex, adding anchors, and deferring sanitization — cutting running time roughly in half. This was a significant improvement but did not reference closing this issue.
  • PR Faster parsing of GC(id) #463 (still open) proposes faster GC(id) parsing by replacing regex with manual string parsing — another incremental performance improvement targeting the same root cause.
  • The parser still uses sequential regex matching across 20+ rules per log line (e.g., 24 rules for ZGC alone). The core architecture of linear rule scanning remains unchanged.
  • Only one commit has touched the parser/ module since June 2023 (a workflow fix), meaning the MRU optimization and decorator improvements are the primary performance changes to date.
  • No duplicate or superseding issues were found covering the same broad performance concern.

• Recommendation: ✅ Keep open — While substantial progress has been made (~28% from MRU ordering + ~50% from decorator optimization), the issue represents an ongoing performance improvement goal that has not been fully resolved. There is still active work (PR Faster parsing of GC(id) #463) targeting this problem, and the fundamental approach of sequential regex matching across many rules per line remains a potential bottleneck for large log files.

Generated by Relevance Check Agent for issue #203