On Fri, May 6, 2005 9:32 am, Peter Farrow said:
Hi There, It seems the default installer with Centos 4, enables DACs and the extra ACLs which DUMP cannot handle.... so I get errors like "DUMP: ACLs in inode #nnnn won't be dumped" since DUMP and RESTORE is integral to my back policy for some 30+ machines I would like to know if its possible to turn off the extra ACLs in the file system as I really don't need them... Can tune2fs do this for me or do I have to re-build the filesystems with some option that I currently don't know...
For Ext3, I have switched to Jorg's STAR because it has done POSIX ACLs for awhile. But when I want to use dump, I have switched to XFS (and its xfsdump), since 2001. You can install with XFS from the installer.
The nice thing about XFS, like Ext3, is that it hasn't changed in structure since the mid-'90s. Unlike IBM JFS, which was ported from OS/2 instead of AIX (largely because of Project Monterey -- see any full and accurate history on the SCO v. IBM lawsuit for more on the reasons why) and lacked a lot of the UNIX/inode structures, XFS was ported directly and fully from Irix to Linux. Which means that POSIX ACLs, Quota, NFS and other things worked off-the-bat in Linux.
In fact, the SGI XFS team was responsible for a _lot_ of the new VFS interfaces of 2.5.3-onward, largely because they had already ported them from Irix to Linux for XFS support. I.e., a lot of the new features of the Linux 2.6 kernel at the VFS that _all_ filesystems benefit from were from Irix's XFS codebase.
As of Linux 2.6 (and late 2.4.25+ kernels?), all of the core support structures that XFS needs are now standard. Other than one bug in XFS 1.0 (fixed some 3 years ago, resulting in the XFS 1.1 release), XFS has been rock-solid and reliable for me. The only major thing that had to change from Irix to Linux was the default block size (to match the default paging of the x86 architecture from MIPS).
-- Bryan J. Smith mailto:b.j.smith@ieee.org