On Wed, Feb 26, 2003 at 03:34:15PM -0500, John Stoffel wrote: > >> The application reported errors like so: > >> > >> info: rendering frame 5 > >> info: frame 5 rendered in 30.485s > >> info: rendering frame 6 > >> error: //N/N/fred1/stress/out.0005.cin: Invalid argument > > Andrew> Note: At this point, there'd be a partially-written frame on the > Andrew> server. > > Well, the obvious questions are: I can answer a few of these now; hopefully all of them later. > 1. How big is the file, The partially written files are some (apparently random) size less than 49186816 bytes. > and how is it corrupted? I suspect it is merely truncated, though I am not sure. I will check this. > Does Shake keep the file around, or does it delete it when > done? It leaves the partially-written file in place. > 2. Have you talked with the Shake developers to get them to > tell you what the errors are or could be? No; I don't believe that we are currently under support. > To me it looks to be an application error of some sort, but > obviously it could be anything. > > 3. Are there any messages on the Samba server side logs? I will check this. > 4. How big are the files on average, They are all precisely 49186816 bytes. > and how are they stored in terms of directory structure? > Are they all in one directory, or does each client get it's > own directory? All in one big directory. > 5. Can you repeat the error by just having some program on the > NT box (such a windows explorer) copy an equivilent number > of same sized files to the Sambad server and see what > happens? Let's try to get Shake out of the equation here > if we can. That's exactly what I was thinking of trying next (since having to come in on the weekend to use Shake - we only have 4 licenses - is a little irritating.) > 6. What version of Samba are you running on the server? 2.2.1a-4. > Basically, I think we need a bunch more info before we can > figure out anything. I personally lean towards an > application/samba interaction, but I could be completely > wrong. That's entirely possible; however, using ext3 rather than XFS causes the problems to become apparent. Andrew Klaassen _______________________________________________ Ext3-users@redhat.com https://listman.redhat.com/mailman/listinfo/ext3-users
