deford@linus.UUCP (Kevin Deford) (03/08/88)
I have just brought up GNU 18.50 on an alliant running OS CONCENTRIX 3.1. It is running NFS and all user home directories are NFS links to a sun server. After compiling and installing GNU, I noticed that everytime I wanted to save a file, I got "IO error writing <filename>". Upon further investigation, I found that when editing files which were on local disk structures, everything was fine. The problem was when editing a file on a remote host - all user directories. More interesting was the fact that it was a semi-intermittant problem. It seemed that if one edited a file and tried to save it, the above error occured. If next the user exited emacs and answered yes to the question "Save file /guru/deford/.cshrc? (y or n)", the changes did get stored correctly. If one then edited that file again, everything was ok, until the user read in another file - at which point the problem re-occured. It seems to be related to the backup files that are dropped out by GNU and then read back in across the NFS link. We have a problem similar to this when dealing with binaries that are located on a NFS structure but are compiled on the Alliant. All binaries must reside on a local disk in order for the kernal to accept them (we get the following rejection: Kernel refused execution of abin/mg.abin. In anycase, has anyone used GNU on an Alliant in a similar situation and had similar problems? -- Kevin M. DeFord Internet: deford@linus%mitre-bedford.arpa; UUCP: linus!mitre-bedford!deford Office: 1A420, phone (617)271-3424 Mitre Corporation, Bedford, MA 01730
pete@esosun.UUCP (Pete Ware) (03/08/88)
1. The "Kernel refused execution ..." message is because Concentrix cannot execute programs residing on an NFS filesystem. This is probably not related to your problem. 2. Since everything works on a local filesystem, I'd suspect NFS. If you have the partitions from the Sun "soft" mounted, then NFS is loosing your file as it is writing it out (a performance problem on the Alliant side). You should have the partitions "hard" mounted -- it will save you months worth of grief, even if you do have to wait every once and a while for a machine to reboot. (See the sun mount(8) man page). The Alliant console should be displaying "NFS server not responding" "NFS server ok" messages when this happens. 3. Maybe it really is an emacs problem. pete -- esosun!pete@seismo.CSS.GOV (Pete Ware) (619) 458-2520 {seismo,sdcsvax}!esosun!pete