Hacktoberfest 2026: los issues que los mantenedores marcaron para octubre, abiertos y aptos para principiantes. Explorar issues de Hacktoberfest

[Bug] Mirrors crash during WAL replay when wal_compression is pglz/lz4

Abierto
  • ¿Algo no está bien?
#2,137 0 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 3 días

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
3/5
Tiempo estimado
Medio día
Aptitud para principiantes
58/100
Tipo de issue
Error
Claridad
Bien especificado
Estado de actividad
Activo
Stack tecnológico
c, postgresql, sql
Área
databases, distributed-systems

Línea de trabajo

Start in RestoreBlockImage() in src/backend/access/transam/xlogreader.c (around line 2096), where zstd_decompress_backupblock() runs before the bimg_info method check. Compare it with XLogCompressBackupBlock() on the write side to see which methods write images. Done when pglz and lz4 images replay without error; reproduce with the SQL in the issue on a cluster with mirrors.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

type: Bug
Apache Cloudberry version

No response

What happened

Reproduce (on a cluster with mirrors):

  CREATE TABLE w (a int, b text) DISTRIBUTED BY (a);
  INSERT INTO w SELECT i, repeat('x', 100) FROM generate_series(1, 1000) i;
  CHECKPOINT;
  SET wal_compression = pglz;
  UPDATE w SET b = 'y' WHERE a < 100;

Actual: the UPDATE succeeds, but all segment mirrors stop WAL replay and shut down, and FTS marks them down:
FATAL: invalid compressed image at 0/C7CFF070, block 0 (error computing decompression size)
LOG: startup process (PID ...) exited with exit code 1
LOG: shutting down due to startup process failure
gprecoverseg (incremental) reports success, but the mirrors replay the same WAL record and crash again. Only gprecoverseg -F (full recovery) brings them back.

Expected: the mirrors replay the WAL normally, as they do with SET wal_compression = zstd. With zstd, the same steps produce no errors.

Cause: in RestoreBlockImage() (src/backend/access/transam/xlogreader.c, around line 2096), every compressed page image is passed to zstd_decompress_backupblock() before the code checks bimg_info for the method
actually used. That zstd-only call dates from commit 60915ec6b2e ("Use zstd for wal_compression", 2020), when Greenplum always compressed WAL with zstd. Since the merges from PostgreSQL,
XLogCompressBackupBlock() writes images with the method chosen by wal_compression (pglz, lz4 or zstd), so pglz and lz4 images reach the zstd decoder and fail. A likely fix is to remove the unconditional zstd
call and rely on the per-method code that follows it.

Impact:

  • A session-level SET wal_compression plus an ordinary write is enough. It leaves the cluster without mirror protection until an administrator runs a full recovery.
  • Not verified: lz4 (expected to fail the same way, since the same code runs first), and whether a primary's own crash recovery also fails after replaying such a record.
What you think should happen instead

No response

How to reproduce

Reproduce (on a cluster with mirrors):

  CREATE TABLE w (a int, b text) DISTRIBUTED BY (a);
  INSERT INTO w SELECT i, repeat('x', 100) FROM generate_series(1, 1000) i;
  CHECKPOINT;
  SET wal_compression = pglz;
  UPDATE w SET b = 'y' WHERE a < 100;
Operating System

any

Anything else

No response

Are you willing to submit PR?
  • Yes, I am willing to submit a PR!
Code of Conduct
Lenguaje dominante
C
Estrellas
1.4k
Forks
262
Merge medio
4 d 20 h
PR fusionados (30 d)
57

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de apache/cloudberry

Todos los issues de apache/cloudberry

Issues similares

Más issues de C

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.