Fsyncgate: errors on fsync are unrecovarable
Read OriginalThis article archives the 'fsyncgate' email thread where Craig Ringer details a critical PostgreSQL flaw. The issue involves PostgreSQL incorrectly handling fsync() EIO errors, particularly on XFS, by retrying failed syncs. This can lead to completed checkpoints despite unwritten data, resulting in potential data loss. The post argues PostgreSQL should PANIC on fsync failures to ensure data safety.
Comments
No comments yet
Be the first to share your thoughts!
Browser Extension
Get instant access to AllDevBlogs from your browser