Ubuntu :: Rsync Between Windows XP And Headless 9.10

I've been trying to get my headless Ubuntu 9.10 server to back up files from my Windows XP box and onto a 1 TB Seagate FreeAgent Go USB drive (which is connected to the Ubuntu server). I've tried two different methods, both of which are behaving strangely and not quite working. I'm using SSH to access my Ubuntu server.

The /etc/fstab line for the USB drive looks like this:

Code:

/dev/sdb1 /media/usb-backup-windows ntfs uid=34,gid=34,umask=022,dirsync,sync 0 0

I'm using NTFS because I'd like to share this drive between Windows and Linux. The drive gets mounted fine and I can read/write files after booting up. The way I'd like for this to work is to have my Ubuntu box mount the Windows drive using CIFS. So I mounted the C drive using the following command:

Code:

smbmount //windowsxp/C$ /mnt/windowsxp/C -o directio,iocharset=utf8,noperm,nounix,ro,credentials=/home/user/.smb/passwords.conf

The mount works fine. I can browse directories under /mnt/windows/xp/C, read files, copy files to Ubuntu, etc. So now I have my USB drive mounted and the C drive on my Windows box mounted. Should be good to go, right? Unfortunately, after several minutes (this varies, sometimes it can go an hour or so) of copying files using rsync --archive /mnt/windowsxp/C/ /media/usb-backup-windows/C/ (the actual command I use has more options - not sure if that's important), the server locks up.

The SSH session dies and I can no longer ping it. The server will eventually start responding after several minutes, only to lock up again a few minutes later, and so on and so on. When it locks up, the following messages end up in /var/log/kern.log:

Code:

CIFS VFS: Unexpected lookup error -26
CIFS VFS: No response to cmd 46 mid 59789
CIFS VFS: Send error in read = -11
CIFS VFS: server not responding

I did some Googling on these messages and came across a suggestion to set /proc/fs/cifs/OplockEnabled to 0. I gave that a shot but it didn't make a difference. I also tried plugging in a mouse and noticed that I could make the server respond immediately after a lock up by moving the mouse. I have to move the mouse though - just leaving it plugged in without movement doesn't help. I have to wait for the hang to occur and then move it. Once I do that, things progress for another few minutes.

This got me thinking that I had a lack of entropy and the mouse movement was kicking things into gear. So I tried moving /dev/random to /dev/random-chaos, and created a symlink /dev/random that just pointed to /dev/urandom. This didn't work - same exact behavior. So why in the world does moving the mouse bring the server back and cause it to start responding, if only for a few more minutes until the next hang?

I then gave up on this approach and tried connecting to an rsync daemon running on my Windows box (using Cygwin), instead of using the CIFS mount point. After getting the config file right and figuring out how to run it as a service on Windows, I started getting files copied once again. However, after what seems to be about the same length of time (several minutes to an hour or so), the rsync connection dies and I get the following message in the Windows rsync log file:

Code:

2010/03/22 13:01:01 [4024] rsync: writefd_unbuffered failed to write 4 bytes to socket [sender]: Connection reset by peer (104)
2010/03/22 13:01:01 [4024] rsync error: error in rsync protocol data stream (code 12) at io.c(1539) [sender=3.0.7]

The Windows box is running rsync 3.0.7 and Ubuntu is running 3.0.6, but they are both protocol 30. The rsync error log on Ubuntu doesn't help much - it also says "Connection reset by peer". I've tried this at least a dozen times and it always fails with these messages. It's weird because it's always 4 bytes, never anything different. I also noticed that /var/log/kern.log had the following messages, although they do not line up with the times that rsync died:

Code:

usb 1-5: reset high speed USB device using ehci_hcd and address 2

I did some Googling on this message and tried some stuff that worked for other people. I added dirsync and sync to /etc/fstab. I tried setting /sys/block/sdb/device/max_sectors to 64. Neither one of those made a difference. unloading the ehci_hcd module and dropping back to USB 1.0. However, 9.10 doesn't seem to have that module loaded, so I'm not exactly sure how to turn off USB 2.0 and just try 1.0. I'm not real enthused with that workaround anyway because I have at least 500 GB to copy.

I've kind of run out of ideas here. It's frustrating because the entire reason I bought the hardware and set up Ubuntu was to run backups. I'm not sure if my problem is a networking issue (CIFS VFS server not responding, connection reset by peer), a problem with running headless (wiggling the mouse temporarily prevents the hang), a USB device problem (reset high speed USB device messages), or something else entirely.