Hey, all. I just said "dump -1au -h 0 -f - / | gzip -c1 > /foo_on_nfs/bar" on a Debian Sarge/i386 box, with foo_on_nfs being a mount from another Sarge/i386 box, and with both / and foo_on_nfs being natively ext3. Dump died with "File size limit exceeded". The file is 2147483647 bytes long, which is a byte short of two gigs. Is there a dump(8) available that does not choke off once its output file size exceeds two gigabytes / 31 bits? This seems to be a pretty serious limitation given the sorts of data one might expect to back up nowadays... I'm pretty sure this limitation is specific to dump(8) on Debian, as I have NetBSD systems dumping much larger files onto the same partition. Thanks in advance for suggestions! I'd really like to avoid the use of tar for this job, but I'll use it if I must, assuming it has no equivalent file size limitations. Gaak! I just noticed something else that's seriously disturbing. This is a NetBSD 2.0 system mounting this particular Debian-served NFS filesystem: acheron /root# ls -l /usr/backup/full/ total 15739508 -rw------- 1 root wheel 7899232767 Feb 20 04:40 20050220-0405-0-5-SLASH.gz -rw------- 1 root wheel 7947841483 Feb 27 04:41 20050227-0405-0-6-SLASH.gz -rw------- 1 root wheel 254421086 Feb 26 18:53 G-20050226-1851-0-6-SLASH.gz Here's a Sarge system with the same filesystem mounted: goliath /root# ls -l /usr/backup/full/ total 15739508 -rw------- 1 root root 3604265471 Feb 20 04:40 20050220-0405-0-5-SLASH.gz -rw------- 1 root root 3652874187 Feb 27 04:41 20050227-0405-0-6-SLASH.gz -rw------- 1 root root 254421086 Feb 26 18:53 G-20050226-1851-0-6-SLASH.gz Here's the view from the server itself: memory /root# ls -l /usr/backup/full/ total 15739508 -rw------- 1 root root 7899232767 Feb 20 04:40 20050220-0405-0-5-SLASH.gz -rw------- 1 root root 7947841483 Feb 27 04:41 20050227-0405-0-6-SLASH.gz -rw------- 1 root root 254421086 Feb 26 18:53 G-20050226-1851-0-6-SLASH.gz There is something very wrong here... Is Debian's NFS client broken for large files? PS: Both Debian systems noted here are running stock kernels. Goliath is running 2.6.8-2-686-smp and Memory is running 2.6.8-1-386. PPS: If I serve the large files listed above from a NetBSD host, both Debian clients are able to see the correct sizes. So, the NFS glitch only appears when a Debian NFS server is being used by a Debian NFS client. Copying the file ends up copying over the size listed by ls, or a bit less than half the actual file. Thanks in advance for clues, advice, and help! -- Mason Loring Bliss (( "In the drowsy dark cave of the mind dreams mason@blisses.org )) build their nest with fragments dropped http://blisses.org/ (( from day's caravan." - Rabindranath Tagore
Attachment:
pgp6_bgF0KGfB.pgp
Description: PGP signature