2.0.25 Aiee via nfs

Greg Louis (glouis@dynamo.dynamicro.on.ca)
16 Nov 1996 12:41:48 GMT

Pulled a stupid move yesterday; consequences were worse than they should
have been, IMHO:

Stupid move was actually two dumb things:

1. Machine B has a mounted directory from machine C but machine C's nfs
daemon has been shut down. Command umount /mnt/C on machine B hangs and has
to be aborted with control-C, after which mount still shows the C dir as
mounted. (Seems to me that in the old (1.2.13) days, umount would work if
this situation arose; unfortunately, I can't pin down the moment it changed;
although I've run most of the kernels since 2.0, I don't make that mistake
very often ;)

2. Machine A has the root of machine B mounted; on A, user root cd's to /B
and does tar -cv --exclude proc --exclude /mnt (sic) -f /dev/nst0 . which
runs along and eventually, because of the spurious / in the --exclude
parameter, tries to back up B's /mnt tree. When it gets to the missing
directory on C tar hangs and has to be stopped with kill -9. Tar: release:
kernel stack corrupted. Aiee. That, of course, was on machine A. On
machine B, nfs is down; existing connections to B don't work ("... still
trying") and new ones can't be made. Machine B is rebooted and all's well.

Both A and B run 2.0.25.

| G r e g  L o u i s                  | pgp:email key@PublicKey.com |
| http://www.consultronics.on.ca/~bgl | id glouis@consultronics.com |
|   Is intolerance of intolerance     |     77 0B 39 81 28 CE 7D D0 |
|   a desirable form of tolerance?    |     4C 7A 6E CC E7 21 4C 55 |