> Jesse Guardiani wrote: > > On Tuesday 02 November 2004 1:55 pm, Casey Zacek wrote: > > > > I do *NOT*, under any circumstances, recommend reiserfs on a > > production machine. One good system crash/power outage and > you can kiss > > your data good-bye. I know from personal experience. Reiserfs4 *may* > > change all of that. We'll see. But until then I highly > recommend EXT3. > > Instead of ext3, I recommend jfs. Its supported in 2.4, the tools install easily in all major distro's, its got a much better on disk layout than ext3, scales better, and is supported by an engineering team with massive experience in file system design and implementation. In my own testing its proven faster than ext3 by a wide margin, and than reiser everytime I've tested on my own workloads. > > On the other hand our mail server has gone down hard twice in > the last 3 > months (once the ups blew up and once a tech pulled out the > wrong power > cord) and both times the reiserfs partition came back up with out any > problems (/home). The ext3 partition on the other hand > needed a manual > fsck the first time and a long rebuild on the second (/ and > /var/qmail). I've tried, hard, to make jfs crap out and never seen it happen. I've had reiser blow up bad, and ext3 blow up bad. Though, for really truly bad failures, nothing takes the cake compared to VxFS from Veritas. > > So your mileage may vary, reiser has come a long way since pre 3.6 > Reiser indeed keeps getting better, however they also keep throwing away the bulk of their prior effort and reverting to crap for long periods of time with each new release.
I solved all of my queue scaling problems by simply buying NetApp filers. WAFL just kicks the ass of everything else out there. Just my own 2 cents. Nick