hallt@handel.CS.ColoState.Edu (Tim C. Hall) (03/06/91)
Hi all, I have an extremely aggravating problem. We run backup scripts nightly that gather all modified files into a monster file in cpio format, then transfer that file to a centralized backup host something like this: find / -mtime -5 -print | cpio -ocB > $BACKUPFILE mount -d BACKUP /usr/remote (where BACKUP is a remote directory advertised through RFS) cp $BACKUPFILE /usr/remote Everything works fine until the cp statement is run. About half the time the file copies successfully. The other half of the time it will bomb out saying "cp: Link has been severed". We've tried using cpio instead, but it does exactly the same thing with the error statement: "cpio:write() in bwrite() failed Link has been severed" I've had this happen to me while running the script manually, and the wierd thing is the remote directory is still mounted and accessible after the failure. I've checked the network immediately after such a failure, and it's fine. Is there any way to fix this problem, or to make cp or cpio more fault tolerant so they will retry? Some days this never happens, other days it does this constantly. Any comments or suggestions are more than welcome. Please reply via e-mail to: thall%spacecom.spacenet-gw.af.mil@server.af.mil Thanks in advance. Mr. Tim C. Hall, System Administrator Air Force Space Command
josie@cbnewsk.att.com (jack.y.gross) (03/08/91)
I would guess that you are running RFS over TCP/IP. If you are you should upgrade to WIN/3B 3.2 which is available now. It has a fix that solves RFS connections dying under load. It is available via the AT&T hotline, 800-922-0354. One word of warning, you must upgrade all machines on your network at the same time or you will experience a MAJOR problem on all the machines on your network. This new version also exists for 6386 and it's called WIN/386 3.2. Jack Gross AT&T Bell Laboratories 307 Middletown Lincroft Rd. Lincroft NJ 07738 (201)576-3197
brad@bradley.bradley.edu (Bradley E. Smith) (03/08/91)
In <1991Mar7.183356.23672@cbnewsk.att.com> josie@cbnewsk.att.com (jack.y.gross) writes: >It has a fix that solves RFS connections dying under load. It is >available via the AT&T hotline, 800-922-0354. One word of warning, >you must upgrade all machines on your network at the same time or you >will experience a MAJOR problem on all the machines on your network. >This new version also exists for 6386 and it's called WIN/386 3.2. No you don't.....what you need to set is the tunable COMPAT42 to 1 in the /etc/master.d/tcp file......first versions of 3.2 tcp didn't have this but the new release does.....I it supposed to be mentioned in the release notes (which we never got). -- Bradley Smith Network & Technical Services @ Bradley University, Peoria, IL brad@bradley.edu --- 309-677-2337