[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference pamsrc::decmessageq

Title:NAS Message Queuing Bus
Notice:KITS/DOC, see 4.*; Entering QARs, see 9.1; Register in 10
Moderator:PAMSRC::MARCUSEN
Created:Wed Feb 27 1991
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2898
Total number of notes:12363

2802.0. "ld, operation fail to complete" by TPOVC::GARYCHENG () Thu Mar 06 1997 06:03

    My customer running DMQ 3.0 on SUN and DMQ 3.1 on WinNT 3.51
    I config NT as group 610 and SUN as group 611 on the same bus.
    After I start the bcp on SUN, the following was spotted in the log:
    
       ld,link listener for group 611 is running
       ld,link sender for group 611 to group 610 is running
       ld,operation fail to complete
    
    From NT side, I can see a temporary connect between NT and SUN, then
    the connection lost.
    Why it connected then lost?
    
    Gary
T.RTitleUserPersonal
Name
DateLines
2802.1xgroup verify? multiple interfaces?WHOS01::ELKINDSteve Elkind, Digital SI @WHOThu Mar 06 1997 11:2635
    Is cross-group verification on?  Does the Sun or the NT machine have
    more than one network interface?
    
    I often see this when there are multiple network interfaces in use, and
    xgroup verification is on.  What typically happens is that the
    initiator ("A") sends its request through one interface on the
    non-initiator ("B"), the one that the host name in A's xgroup table
    resolves to - then the outbound link is established.  Then B sends its
    request for the inbound link to A - but B's routing table is configured
    to send to A's IP address through the second interface whose IP address
    does not match the hostname in A's table.  Right host name, wrong IP. 
    You can see this on A by using netstat -a during the brief time when
    the inbound link exists, but at least in v3.2A, you see no log entry on
    A other than "ld, operation failed to complete" and the exit messages
    for dmqld (at least it doesn't when B's second address is not
    resolvable to a name).
    
    Unix DmQ v3.x, with the exclusion of v3.2A, does not support operation
    over multiple interfaces.  If you do have multiple interfaces, and one
    or more that do not match the IP address of "hostname" is being used
    for DmQ traffic, then you need to use v3.2A on that machine - or turn
    xgroup verify off on the other machine(s).
    
    My guess is that the same applies to NT DmQ, since it's based on the
    same code, but DmQ engineering will have to answer that one.  Also,
    there is no 3.2A for NT, I guess you will may have to wait for 4.0 on
    NT for multiple-interface support (ask engineering on that one too).
    
    (In the case of B using something like ServiceGuard or Sun IP
    Switchover (or whatever it's called), with "floating IP" addresses, B's
    host name will not match either, and you will get a log message on A
    saying "host XYZ port 11111 not found in local address database" or
    something similar, where XYZ is the "native" hostname, which was not in
    A's xgroup table (i.e., XYZ is what you get when you use the "hostname"
    command on B)).