Patrick J. LoPresti
2010-Jun-12 00:54 UTC
[Ocfs2-users] OCFS2 and huge (> 50TB) partitions
Hello. I am experimenting with OCFS2 on a brand new 10GigE iSCSI SAN. It looks pretty good so far -- I am seeing sustained (direct I/O) read speeds of 1300+ MB/sec -- except for one problem: I need a single partition to hold much more than 16 TB. I am running into the same issue described by this thread: http://www.mail-archive.com/ocfs2-devel at oss.oracle.com/msg04740.html That is, even though I created a partition with "mkfs.ocfs2 -C 1M -J block64 --fs-feature-level=max-features..." and mounted it with "-o inode64", it still fails with: "Volume might try to write to blocks beyond what jbd can address in 32 bits.">From the description on that thread, this is an obsolete check infs/ocfs2/super.c. But when I look at the kernel source for 2.6.32.15 and 2.5.35-rc2, I see that the check is still there. I believe this means that partitions larger than 16TB will not work on any released kernel. Is my understanding correct that this check prevents large partitions from working at all? Do the OCFS2 developers intend to remove it? When? I am doing my testing on Suse Linux Enterprise 11 Service Pack 1, which is based on kernel 2.6.32 and ocfs2-tools 1.4.3. My research suggested these software levels should support large partitions. I do intend to take this up with Suse, but I was hoping there was an official patch I could point them to... Thanks! - Pat
We could remove this check. If you want this in your sles kernel, the quickest route will be via Novell. We'll need both sles and mainline patched. On Jun 11, 2010, at 5:54 PM, "Patrick J. LoPresti" <lopresti at gmail.com> wrote:> Hello. I am experimenting with OCFS2 on a brand new 10GigE iSCSI SAN. > It looks pretty good so far -- I am seeing sustained (direct I/O) > read speeds of 1300+ MB/sec -- except for one problem: I need a > single partition to hold much more than 16 TB. > > I am running into the same issue described by this thread: > > http://www.mail-archive.com/ocfs2-devel at oss.oracle.com/msg04740.html > > That is, even though I created a partition with "mkfs.ocfs2 -C 1M -J > block64 --fs-feature-level=max-features..." and mounted it with "-o > inode64", it still fails with: > > "Volume might try to write to blocks beyond what jbd can address in > 32 bits." > >> From the description on that thread, this is an obsolete check in > fs/ocfs2/super.c. But when I look at the kernel source for 2.6.32.15 > and 2.5.35-rc2, I see that the check is still there. I believe this > means that partitions larger than 16TB will not work on any released > kernel. > > Is my understanding correct that this check prevents large partitions > from working at all? Do the OCFS2 developers intend to remove it? > When? > > I am doing my testing on Suse Linux Enterprise 11 Service Pack 1, > which is based on kernel 2.6.32 and ocfs2-tools 1.4.3. My research > suggested these software levels should support large partitions. I do > intend to take this up with Suse, but I was hoping there was an > official patch I could point them to... > > Thanks! > > - Pat > > _______________________________________________ > Ocfs2-users mailing list > Ocfs2-users at oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-users