Date: Wed, 26 Jan 2005 15:53:46 -0700 (MST) From: "Terry R. Friedrichsen" <terry@uplift.hosp.misyshealthcare.com> To: freebsd-questions@freebsd.org Subject: re: FreeBSD 5.3 file system troubles Message-ID: <200501262253.j0QMrkOJ043661@uplift.hosp.misyshealthcare.com>
next in thread | raw e-mail | index | archive | help
I wrote: >> try running something that works the file system and simply turn off the >> system power in the middle of it. to which kris@obsecurity.org replied: > This is expected if you don't turn off write caching of the hard > disks. It breaks the softupdates consistency model because data > written to the disk may not actually be written to the disk, so it's > not there following an unexpected power cycle. Unfortunately write > caching causes a performance hit, and there was a large user backlash > when it was briefly enabled by default some years ago. What you say is true, but what I'm observing is far worse than simply missing the last few blocks of output files, etc. The last time I had to power-cycle the Alpha box (because Xorg hung it), it rebooted to single-user mode but I couldn't even run "sh" because some file in lib was missing. Or if I *do* get into "sh" to run fsck, it finds *hundreds and hundreds* of problems ... And all of this is on a freshly-installed, bog-standard 5.3 system. Anyway, I'm going to stop messing about with Xorg on that machine, which will doubtless make the problem invisible. And yeah, I know I'm gonna have to stop upgrading my Alpha machines some day, but I was hoping to get 5.something with an X system running as an end-of-life position. The i386 SMP box, though, is another story. I am going to have to nail down its problems if I intend to track FreeBSD on it. If I could suc- cessfully build a kernel on it, I'd turn off SMP and see how it behaves. But it appears that I am the only one suffering ... Thanks for all the responses. Terry terry@uplift.hosp.misyshealthcare.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200501262253.j0QMrkOJ043661>