Moved all files off the raid, ran reiserfsck with --rebuild-tree (fix-fixable was not enough), ran rsync after rebuild finished -- no problems.
There were a number of files that the rebuild-tree found that weren't attached
Edward King wrote:
After switching much hardware (and getting some helpful suggestions) I moved the specific machine's files on the backup server to a hard drive outside the raid (still on the backup server, /dev/hdi1) and tried rsync -- problem solved.
It seems there's a problem with the journaled filesystems (reiserfs) and raid -- reading the files is ok (cp command worked fine), writing them is not.
I'm going to turn on reiser debugging and internal checks and re-run -- maybe I can send something to the reiser or kernel people of interest (don't want to go back to ext2)
jw schultz wrote:
On Fri, Jan 10, 2003 at 02:25:57PM -0600, Edward King wrote:Has anyone seen this? Looking for past experiences / ideas. Will post progress.I'm tracking down a problem that seems to be caused by rsync. When moving files from a remote server I get a kernel panic. We have a number of servers that back up to a main box -- the panic only occurs when a specific client backs up. It occurs on the box it is backing up to -- not the client. The kernel is Linux 2.4.20 (just compiled -- no patches), files are stored on a 4 disk raid (80GB Western Digital drives, software raid) with Reiserfs. This is being done over a vpn connection controlled by another machine (gateway machine) running tinc so we're not using ssh or any other shell on the rsync machine. I have: recompiled rsync at both locations recompiled the kernel from new source code (panic in 2.4.19, system rebooted in 2.4.20) I will: try different hardware run a file system check at the main system (the one that crashes) exclude directories in the backup (rsync one directory at a time -- see where it crashes) I did notice filenames on the client machine that contain control characters -- but they seem to have backed up before.Just to confirm: You are doing backup-server initiated pull something like "rsync server::module destdir" and the machine you execute this on (the receiver) panics. Rsync will of course not be the culprit. However, rsync is very good at stressing a system and the kernel developers have found it to be a common test case. Most likely it is a hardware fault. Probably timing sensitive. I'd check the logs for oopses and and disk errors. Also try downgrading the mode with hdparm. If this isn't a obvious hardware fault you should report it to the linux-kernel people. Look on www.kernelnewbies.org for instructions. This will be of interest to the developers of md, reiserfs and the specific IDE driver.