crampton@hpldsla.sid.hp.com (David Crampton) (02/01/91)
I am asking for help in determining the cause of failure of file transfers using 'rcp'. Running HP-Ux 7.03 on an HP9000/340 to as many as four HP Vectras (PC clones) running FTP/PC version 2.04 programs including 3c500 driver and 'rcp'. Each of the four PCs invokes rcp to download a complex dir structure, eg, rcp -r <pathname/*> ; the same dir structure to each PC. Each PC has a different IP address of course. Up to 2 PCs running this process has worked perfectly for over 1 yr. Expanding to 4-at-once causes all four PCs to receive only about 1/2 of the fileset. The files transferred are correct but files are missing, eg. 84 files should be 155. All four PCs receive what appears to be the same truncated fileset. The rcp errorlevel is NOT set to 1 so my DOS batch file zooms on through to some error caused by a missing file. My own idea is that a resource limitation on the Unix side is killing the rcp processes. The pathname in the rcp call is fairly long (about 25 chars, 8 levels down). File /etc/conf/dfile contains no special network tuning statements. The netlogstart process is NOT posting errors to neterrlog; but it may have been started with an incorrect combination of options, ie. ( no -C where -f exists) . Quite annoying of course that rcp stops prematurely and I'm not getting an error signal. I will appreciate help or pointers to. Would like to understand the process before I go randomly rebuilding the kernel. Many thanks. -------------- David Crampton Manufacturing Engineering, Software HP Scientific Instruments Div, Palo Alto Ca.