1. 6

  2. 4

    I don’t like how this is branded online specifically as a PostgreSQL failure. This was an issue caused by developers having wrong assumptions about fsync behavior. In fact it differs per OS and even per OS version on Linux. This issue by far impacts much more than just PostgreSQL - every application that expects that fsync is a reliable way to dump data to storage is now suspect to data loss until reviewed. PostgreSQL just happened to step on the mine first but that doesn’t change the fact that a platoon is still in the middle of a mine field.