wrat@unisql.UUCP (Louis Marco) (06/07/91)
(First of all my thanks to all who responded to me on my gdb problems; email's still an iffy thing here so I may not ever reply directly, but thanks for your help. Now for the next question... :-) We have a couple of dozen Suns and a single 386 running Interactive 3.2. I have a file system physically on the Sun that's the file server mounted via nfs on the 386 box. Sometimes (not all the time, but often enough so that I can't work the way I want) when I do an I/O intensive operation such that the program is executing on the 386 but the data is read from the Sun, the console on the (Sun) file server fills up with rfs_read: attempt to read from non-file messages and the load averages on the server skyrocket. What's particularly weird is we don't have rfs installed anywhere. (The string is from nfs_server.o, part of the Sun vmunix). The problem only appears in Sun/386 interactions. If I run make on the source in the nfs-mounted directory, or copy a large file from the nfs-mounted directory I eventually get the above. Why, and what should I do to make it stop? wr
ir@crosfield.co.uk (ian reid) (06/07/91)
In article <1334@unisql.UUCP> wrat@unisql.UUCP (Louis Marco) writes: > > We have a couple of dozen Suns and a single 386 running Interactive >3.2. I have a file system physically on the Sun that's the file server >mounted via nfs on the 386 box. Sometimes (not all the time, but often >enough so that I can't work the way I want) when I do an I/O intensive >operation such that the program is executing on the 386 but the data is >read from the Sun, the console on the (Sun) file server fills up with > >rfs_read: attempt to read from non-file We had a similar problem at our site when we were running ISC 2.0.1 and NFS 2.0 using a Sun 3/??? running SunOS 3.5 (in the days before it was called SunOS). One thing that seemed to cause the problem was if the shell invoked to process non binary scripts (set by the built-in variable shell in the C shell, or by the environment variable SHELL for the Bourne shell) was the C shell. Changing this to be the Bourne Shell reduced these errors greatly, but never totally eradicated them (although it's not clear if everyone did use this workaround). I can't tell you if ISC 2.2 and NFS 2.1 have the same problems because at the time we switched to 2.2 we also started using a PC box as a fileserver, and this problem only ever occurred between a PC and a Sun. Hope this helps! -- Ian Reid #include <std/disclaimer.h> UUCP: ir@crosfield.co.uk or ...!{ukc,mcsun,uunet}!cel!ir "Microsoft Windows - Gates way to the future. I'll put my X against that."