[Bug] CheckpointFile skips valid backup recovery when the primary numeric header is malformed

オープン
  • 何か問題がありますか？
#11,460 コメント 1 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

@hickey-ai がすでに取り組んでいます。

2026年10月20日 から。

  • #11463 @hickey-ai による — オープン

評価

難易度
3/5
見積もり時間
1〜2日
初心者へのやさしさ
35/100
issue の種類
バグ
明瞭さ
明確に書かれている
活発さ
停滞
技術スタック
java
領域
backend

調査の方向性

Start by reading CheckpointFile.read(String) and the public read() recovery path; also inspect EpochFileCache.initCacheFromFile and the existing remoting CheckpointFileTest tests. Reproduce with a malformed primary numeric header and a valid .bak, then verify recovery, failure cause and path when both headers are invalid, and regression behavior for valid primaries and CRC/count mismatches. PR #11032 is unrelated; open PR #11063 already proposes this fix.

索引モデルが issue の本文から書いたものです。

説明

Before Creating the Bug Report

  • I found a bug, not just asking a question, which should be created in GitHub Discussions.

  • I have searched the GitHub Issues and GitHub Discussions of this repository and believe that this is not a duplicate.

  • I have confirmed that this bug belongs to the current repository, not other repositories of RocketMQ.

Runtime platform environment

Windows 11; local checkpoint files and native common/remoting module tests. No live broker or HA failover claim.

RocketMQ version

develop, revision 5.5.1, commit 57c2dc04f78cc7e73e6b8c9436ee0fff0d0f46c3.

JDK Version

OpenJDK 1.8.0_302 (GraalVM CE 21.2.0), Maven 3.6.3.

Describe the Bug

CheckpointFile.read(String) parses the entry count and CRC with Integer.parseInt. An empty, truncated or nonnumeric header throws NumberFormatException, while public read() only catches IOException to recover from .bak. A valid backup is therefore ignored. EpochFileCache.initCacheFromFile also only catches IOException.

Steps to Reproduce
  1. Create a CheckpointFile whose serializer writes and reads the entry string unchanged.
  2. Write Collections.singletonList("saved") and copy the checkpoint file to checkpoint.getBackFilePath().
  3. Replace the primary file with "invalid\n0\nvalue\n" (or an empty file).
  4. Call checkpoint.read().

The valid backup can be read on its own; the malformed primary prevents reaching it.

What Did You Expect to See?

Malformed numeric headers enter the existing IOException recovery path and the valid backup is returned. If the backup header is also invalid, the final failure is IOException with the failing path and NumberFormatException as its cause.

What Did You See Instead?

An unchecked NumberFormatException escapes from parsing the primary header instead of loading the valid backup.

Additional Context

The proposed fix only wraps the two numeric header parses; serializer exceptions are not swallowed. Existing valid-primary precedence and CRC/count mismatch behavior are regression controls. Six new header tests plus the three existing remoting CheckpointFileTest tests pass on the independent patch (9 native tests, 0 failures/errors/skips). Baseline native Surefire reproduced the NumberFormatException and stopped after the first failure; not all six new methods ran before the fix. Checkstyle/SpotBugs pass.

PR #11032 concerns empty checkpoint state and does not handle malformed numeric headers. Issue/PR/discussion searches did not identify this header-recovery fix.

AI assistance: OpenAI Codex was used for investigation, implementation and tests.

主要言語
Java
スター
22.7k
フォーク
12k
平均マージ
4日 20時間
マージ済み PR（30日）
11

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

apache/rocketmq のほかの issue

apache/rocketmq の issue をすべて見る

似ている issue

Java の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。