On 9/21/07, UNIX admin <[EMAIL PROTECTED]> wrote: > > XFS? > > XFS! In absence of ZFS, the best 'classic' filesystem out there, bar none, no > ifs, buts, or maybes.
My, my, touchy are we? > > > I do not see the serious error of not turning on > > a filesystem > > that is 1) not going to work properly with 4k stacks > > (for how long > > now?!?!) > > Works perfectly on IRIX. Gee, I wonder why? Now, is it the responsibility of > a filesystem to work around crappy engineering, or lack thereof in the Linux > OS? So please point out the serious error in not including a filesystem that is effectively a penguin or seal out of water and on land? > > If the answer is "yes", something is definitely wrong... Duh...I doubt that one would hold the penguin/seal responsible for being put in a race on land by its owner. Perhaps SGI should have ported IRIX over to x86 and open sourced it. > > > and 2) has not become any more stable or > > reliable than its > > best form in the 2.4.18 to 2.4.22 versions of the > > Linux kernel. So > > what if XFS is the best performing filesystem at > > writes on Linux? If > > you lose power, you are going to get plenty of > > corrupted files if it > > was doing heavy writing when it happens. > > You are writing to the wrong guy about stability. We have XFS running on > CentOS of all things, and we have done extreme stress tests on that combo; > not once did the system crash or did we lose any data. Hey, I can do that too...with the appropriate hardware. RAID with BBU write cache and I will get blazing fast performance and reliabililty with any filesystem on Linux. > > I really don't want to get dragged into an XFS discussion here, but your > statement about losing data is technically plain wrong, since XFS is a > journaling system. I know for a fact that I had people pull the plugs on my > SGI Challenge server while the thing was in the middle of R/W, and not once > did I lose any data. So like I wrote, I'm really the wrong guy for you to be > arguing about XFS stability with. Yeah, wish I had your iron. I have lost thousands of mails because the mail queue was on a XFS filesystem and files would be there but full of @@@@@.... > > > and the fact that to work around the differences from > > IRIX and Linux > > has turned the XFS code into such a monster that > > quite a few Linux > > kernel developers have publicly proclaimed that they > > want nothing to > > do with XFS code. > > Again, that says more about Linux than it does about XFS, and not in a > positive way. I really doubt XFS would have fared much better had it been ported to Solaris or if it is ported to OpenSolaris. It is just not the native environment. > > > [XFS] Fix race in xfs_write() b/w dmapi callout and > > direct I/O checks. > > [XFS] Fix to prevent the notorious 'NULL files' > > problem after a crash. > > [XFS] Fix race condition in xfs_write(). > > [XFS] remove more misc. unused args > > [XFS] reducing the number of random number functions. > > All that means is that XFS is being actively worked on. Good. Cor, race conditions after all this time? Why did it take them so long to find it? > > > XFS on Linux is a mess. > > Again, says more about Linux than XFS. > Let us see you gather a team and port XFS over to OpenSolaris and see how you fare? A penguin/seal out of water is a penguin/seal out of water. No matter how great the penguin/seal is in the water, it becomes a entirely different matter when it is elsewhere. Let us see IRIX on x86 instead. _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
