News for hackers. Stuff that matters.
Re:WEKA Responds to Unfounded Allegations Made by MinIO Rega... by tobias3 on Monday March 27, 2023 @05:49PM (#35333180)
MinIO had a large investment round at unicorn valuation at the end of last year. Watch them desparately move up market (or more like flailing around) to recoup that investment.

Meanwhile they cannot get their software to work on ext4 and it is apparently ext4s fault[0].

[0] https://github.com/minio/minio/issues/16602#issuecomment-142...

Re:WEKA Responds to Unfounded Allegations Made by MinIO Rega... by dekhn on Monday March 27, 2023 @08:42PM (#35334243)
You're misinterpreting the bug, the software works but could lose data. However, the person who wrote the update about O_DIRECT and ext4 in production environments is just wrong.
Re:WEKA Responds to Unfounded Allegations Made by MinIO Rega... by singron on Monday March 27, 2023 @09:21PM (#35334574)
If it loses data, does it work?
Re:WEKA Responds to Unfounded Allegations Made by MinIO Rega... by dekhn on Monday March 27, 2023 @09:34PM (#35334687)
I mean, losing data is an extremely complex problem. What I'll say is that I've run large production systems with heavy load on ext2 and ext4 and did not find that it "lost data" more frequently than xfs, and in fact performed fine.

What's really important here is the person who closed the bug did some summarily, with little to no explanation, other than a reference to O_DIRECT not being supported in ext4, and therefore it "loses data" in production. This is an unprofessional comment to close a bug and I'd expect one of the co-founders of MinIO to do a better job. See https://ext4.wiki.kernel.org/index.php/Clarifying_Direct_IO%... for a discussion of the subletly of O_DIRECT. It's also not really required to make a production filesystem reliable, and would only be a tiny part of an overall reliability solution (because drives themselves often buffer and reorder their writes, sometimes even lying about whether data was "commited durably to disk")