mccalpin@MASIG1.OCEAN.FSU.EDU ("John D. McCalpin") (05/08/88)
I have had repeated trouble with an NFS problem that sounds suspiciously like the "vanishing .rhosts" problem just reported by phil@brl.arpa. I have a 3030 and a 3130, with most of the 3130's /usr partition mounted on the 3030. When working on the 3130 (client), but accessing the 3030 (server) disk, I find that interrupting a program that is writing to a file on the server often causes that file be be made permanently unavailable to the client. The directory listing is fine, but cat returns nothing. I have to re-start the nfs daemons to get access to the file again. The file is perfectly accessible from the server machine side. This is extremely frustrating. Unfortunately, it gets worse.... When I interrupt a Make (for example) in the few fractions of a second during which it is loading an executable (like a compiler) from the server, I find that that file becomes inaccessible as well! When I try to access it again, I get the message "Interrrupted System Call", and again have to re-start the nfs daemons.... Have other people seen this trouble ??? john mccalpin mccalpin@nu.cs.fsu.edu mccalpin@masig1.ocean.fsu.edu P.S. All this is not to mention the fact that I still cannot cp a large file between machines without having it occassionally corrupted. It is pretty silly to tell all the users that they do not need to pay any attention to what machine a file is on, unless it happens to be bigger that about 1MB, in which case they have to use rcp to copy it!!! SGI has told me that this is a known bug with no known fix --- great!....