ben@pka6.dk (bent nielsen) (01/12/90)
Hello World ! HELP !!! We're running Oracle software on 5 AT&T 3B2/600s and 2 Pyramid MI Servers. All of these are connected by backbone network consisting of an ethernet with AT&T's (i.e. Wollongong's) Enhanced TCP/IP WIN/3B software. The 3B2s are also connected by a terminal network consisting of twisted pair cables and STARLAN software. Thus, our configuration looks like this: Pyramid Pyramid ! ! -----------*------*---------*--------*---- TCP/IP WIN/3B ! ! ! ! 3B2 3B2 .... 5 in all ! ! ! ! -----*-----*-------*--------*------*------ STARLAN ! ! ! ! ! ! PC PC PC .... approx. 230 PCs The PCs all use TEWORKS terminal emulators to access the 3B2s. The 3B2s all run the Oracle database software and/or act as front ends for 2 Pyramid MI Servers running the Oracle software using SQL*Net. Now, our problem: There are a number of programs that seemingly can't be run on a 3B2 from a terminal connected via the STARLAN network, without all of their terminal output disappearing into the great unknown. Also, even small (timewise) programs take an inordinately long time to complete. If you run the same software from a terminal connected directly to one of the TTY ports of the 3B2, all is well. Also, if you run the same software on a 3B2 from a terminal connected via the TCP/IP `rlogin' program through another 3B2 or Pyramid, all is well too no matter what kind of connection you have to the other 3B2. If you do run these programs on a 3B2 from a terminal connected via STARLAN, but pipe the output (stderr included) through a filter such as cat, as in remsh <some host> "echo hello; echo again >&2;" 2>&1 | cat or rlogin <some host> 2>&1 | cat then again, all is well. Examples of programs that have this problem: All of the Enhanced TCP/IP WIN/3B programs (rlogin, remsh {i.e. rsh in System V disguise} ftp, rwho, rcp, etc. etc.). Oracle programs that incorporate routines from the TCP/IP network libraries. AT&T programs such as ESCORT from the 3270 Emulator+ package. This has led to the suspicion that the /dev/slan?? pseudodevices do not behave as a TTY device should behave, at least when used in conjunction with TCP/IP WIN/3B software, but what can we do about it ? Another problem concerns the stability of the telnet and rlogin programs. Once in a while the connection breaks, apparently without cause - and often without the display of the 'Connection closed' message. This is quite a pain, since our users have to be distributed evenly amongst the 3B2s (because of limitations in the number of STARLAN connections per 3B2), and a lot of them therefore have to `rlogin' further on to the 3B2 running their application. Needless to say that they're not impressed when their connection drops dead in mid-transaction. We've experienced this problem in all (combinations of) the following versions and releases of the software mentioned above: System V: 3.1.1, 3.2.1 Oracle: 5.1.17.4, 5.1.22, 6.0.24 Enhanced TCP/IP WIN/3B: 2.1 Starlan: 1.1.0 Now, we've tried using the 3.0 release of TCP/IP WIN/3B, and it appears to make everything work as expected. BUT (surprise !) Oracle does not support the use of this release. We've tried asking Olivetti, our supplier of AT&T hardware and software, we've tried asking the Danish Oracle branch, and Olivetti has had people from AT&T flown in from the US to look at it, but all to no avail. So, now we're asking you: Does anyone out there have similar problems ? Does anyone out there have a solution ? Thank you, Bent -------------------------------------------------------------------------- PKA Phone: +45 42 85 15 22 ext. 5104 att: Bent Nielsen Fax: +45 42 85 15 32 Bredevej 2 UUCP: ben@pka6.dk DK 2830 Virum Denmark --------------------------------------------------------------------------