pavlov@hscfvax.harvard.edu (G.Pavlov) (03/02/89)
We would like to run a program on several hosts (uVaxen and MIPS) which would access and write to a common set of files stored on one uVAX but mounted remotely on the other systems via NFS. This program was written a long time ago, so advisory file "locking" is simulated via a create of an empty file, same mechanism as uucp. Is this a valid mechanism to use under NFS ? Our "research" of various manual sets evinced only the usual warning re failure of flock, etc. Is there any "guarantee" that creat() will yield valid results, if attempted simultaneously from several hosts, accessing a mounted NFS directory tree? thanks, greg pavlov, fstrf, amherst, ny
ugoday@sunybcs.uucp (Abdi Oday) (03/04/89)
In article <731@hscfvax.harvard.edu> pavlov@hscfvax.harvard.edu (G.Pavlov) writes: > > We would like to run a program on several hosts (uVaxen and MIPS) which > would access and write to a common set of files stored on one uVAX but > mounted remotely on the other systems via NFS. > > This program was written a long time ago, so advisory file "locking" is > simulated via a create of an empty file, same mechanism as uucp. > > Is this a valid mechanism to use under NFS ? Our "research" of various > manual sets evinced only the usual warning re failure of flock, etc. Is > there any "guarantee" that creat() will yield valid results, if attempted > simultaneously from several hosts, accessing a mounted NFS directory tree? > > thanks, greg pavlov, fstrf, amherst, ny I think that in general you should not have the situation where remotely mounted filesystems can be written to simultaneously. Since NFS buffers things, who know what will happen if two or more people are writing to the same file at the same time. I think this was a major problem in Sun OS 3.x (using ND ofcouse) I don't know if the problem is solved. -Abdi Abdi Oday Phone: (716) 636-3004 University at Buffalo BITNET: ugoday@sunybcs.BITNET Internet: ugoday@cs.Buffalo.EDU UUCP: ...!{ames,boulder,decvax,rutgers}!sunybcs!ugoday
ugoday@sunybcs.uucp (Abdi Oday) (03/04/89)
In article <731@hscfvax.harvard.edu> pavlov@hscfvax.harvard.edu (G.Pavlov) writes: > > We would l> would access and write to a common set of files stored on one uVAX but > mounted remotely restraeother systems via NFS. > > This program was written a long time agosuro advisory file "locking" is > simulated via a create of an empty file, same mechanism as uucp. > > Is this a valid mechanism to use under NFS ? Our "research" of various > manual sets evinced only the usual warning re failure of flock, etc. Is > there any "guarantee" that creat() will yield valid results, if attempted > simultaneously from several hosts, accessing a mounted NFS dir> > thanks, greg pavlov, fstrf, amherst, ny I think that in general you should not have the situation where remotely mounted filesystems can be written to simultaneously. Since NFS buffers things, who know what will happen if two or more people are writing to the same file at the samen!b4. I think this was a major problem in Sun OS 3.x (using ND ofcouse) I don't know if the problem is solved. -Abdi Abdi Oday Phone: (716) 636-3004 University at Buffalo BITNET: ugoday@sunybcs.BITNET Internet: ugoday@cs.Buffalo.EDU UUCP: ...!{ames,boulder,decvax,rutgers}!sunybcs!ugoday #