Conference help::decnet-osi_for_vms

3914.0. "error handling for PATHLOST" by HYDRA::SCHAFER (Mark Schafer, SPE MRO) Wed Apr 02 1997 09:39

    I'm trying to help a partner that's having problems with error
    handling. I'm trying to get more details (including whether this is
    happening on a VAX or on Alpha), but I'd appreciate knowing if what he
    describes is the intended when there are failures.
    
    Mark Schafer
    SPE MRO
    297-3524
-----------

We have migrated our client-server applications from DECnet IV
to DECnet/OSI. It works fine on DECnet/OSI version 6.1, but
not on DECnet/OSI version 6.3.

The software runs on VMS, is written in C, and uses the
VMS QIO API to DECnet, using what DEC calls "non-transparant
task-to-task communication".

The problem is this: when the DECnet connection is physically
broken, a SYSTEM-F-PATHLOST event is signalled. As per the
API documentation, I have both a VMS I/O channel and an
associated VMS Mailbox. I also have an asynchronous QIO READ
operation pending on both, with associated AST callback functions.
When an event like SYSTEM-F-PATHLOST occurs, both AST functions
should be called. The AST for the channel read QIO should have
SYSTEM-F-PATHLOST in the associated IOSB I/O Status Block.
The AST for the mailbox read QIO should read a MSG$_PATHLOST
message.

This happens on DECnet IV, DECnet/OSI 6.1 but *not* on DECnet/OSI
6.3, where the mailbox message never arrives. This confuses our
software, since the DECnet channel now is in an unspecified state.
(I can go into lots of more detail on the latter subject, if needed).

Please help!

Best regards,

Per Lindberg ([email protected])         !      __!__        Ceci n'est pas
Front Capital Systems, Box 70351,   !  _____(_)_____    un Piper
10724 Stockholm Sweden. 46-84540000 !     !  !  !

Problem Statement