Date: Tue, 4 Oct 2016 19:34:11 +0000 From: Colin Percival <cperciva@tarsnap.com> To: freebsd-cloud@freebsd.org, freebsd-stable@freebsd.org Subject: HEADS UP: Do not upgrade EC2 instances from 10.x to 11.x yet Message-ID: <010001579130685e-fc984470-c447-442d-8dde-7992e3501cf0-000000@email.amazonses.com>
next in thread | raw e-mail | index | archive | help
[Apologies if anyone gets this twice; the first copy I sent seems to have been eaten by a mail server somewhere.] We've identified a bug in the loader(8) in 11.0-RELEASE (to be precise, FreeBSD 11 after April 6th) which results in it attempting to read past the end of the disk if the last partition is not aligned to a 4k boundary. On most (maybe all) physical hardware this results in significant delays to the boot while the spurious I/O fails; in Amazon EC2, this results in the instance hanging permanently. Most systems do not have such misaligned partitions, but the FreeBSD 10.x images in EC2 do, and will consequently hang on reboot if you upgrade them to 11.0 (or to 12-current, for that matter). I recommend not doing this. The AMIs which have been built for FreeBSD 11.0 have properly aligned partitions, and are not affected by this bug, so (once the release is out!) you'll still be able to get FreeBSD 11.0 by launching new EC2 instances. I imagine that this will be fixed with an errata notice shortly after the release, after which point it will be safe to upgrade (since you'll end up with the fixed loader), but as always that will ultimately be up to the release engineering team. Thanks to Peter Ankerstål, Allan Jude, Warner Losh, and Glen Barber for their help in tracking down this problem. -- Colin Percival Security Officer Emeritus, FreeBSD | The power to serve Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?010001579130685e-fc984470-c447-442d-8dde-7992e3501cf0-000000>