Conference iosg::all-in-1_v30

The seven character node-name looks suspect to me.  If I assume it is
actually a valid DECnet nodename (I don't know enough about DECnet to check),
it wouldn't surprise me if UA, or the data files it uses, can't support that.

The node field in NETWORK.DAT, for example, is only six characters.

Guessing even more, the "Illegal usage for device" suggests to me that due
to the above length overflow, somewhere you're losing a colon�, and hence trying
to open a DECnet link to (what it thinks is) a local disk, rather than a
remote node.

But then I'm probably wrong....

Scott

�: Eg by writing "POLARIS" into the beginning of a string which has six blanks
then two colons.

    Actually UA refers to the node field in MNODES (SM MM ISM). But it is
    also 6 characters. Just change it to be the node on which ALL-IN-1
    runs. I'm not sure about the node-name in the user's NETWORK record,
    but I think it is better to set it to (again) the node on which
    ALL-IN-1 runs.
    
    Sjaak.

    
    	I've seen this sort of thing if you access a machine whilst it
    	is still in the reboot sequence. OA$DATA will only be defined
    	when A1V30START has actually run, so if you access things before
    	then, you get -RMS-F-DEV, error in device name or inappropriate
    	device type for operation (which isn't exactly what you said, but
    	what you said looks a like a BASIC error message, which could come
    	from this).
    
    Regards,
    Paul

    Sorry about that POLARIS stuff, it should be POLAR, I misheard over
    the phone (and didn't count).  So it's quit legit.
    
    The reboot thing is not on. This happens like every second night,
    and they're not booting then. 
    
    Since the UA runs an EXE file, you can't see what might happen on
    the remote node.... 
    
    So I'm still wondering about the cluster alias thing. In this case
    where they have a cluster but ALL-IN-1 is only installed on one of
    the nodes (i.e. you have to log into that node to run ALL-IN-1),
    should they have the cluster alias or the node name on SM MM ISM?
    
    How about the entries in NETWORK.DAT? Should they have cluster
    alias or node name in the NODE field?  It seems they have the cluster
    alias, if it will have to be changed, what is the easiest way to
    do it and get correct databases on all the nodes?
    
    Thanks,
    Hans

    Hans,
    
    I'd definitely change the Network Profile Addresses and the ISM file to
    have just the individual node name in them.
    (for the users just empty NETWORK and then use the SMNETINIT com file
    to repopulate them).
    
    Also you can run the UA housekeeping job in the old pre-v23 
    (aka simple) way by doing
    
    $RUN OA$LIB:SMNETUPDT
    
    and see whether that works it does give some good trace info that is
    unfortunately hidden in the log produced when UA is run as a HK job.
    
    regards,
    
    Andrew.D.Wicks
                                                          

    Thanks Andrew, 
    
    just one more on the node thing, I notice that SMNETINIT uses 
    OA$PRIMARY_NODE, but with SM MM MAD C you get OA$NODE in the
    node field. 
    
    So what is really OA$PRIMARY_NODE? I see somewhere it's defined
    as "primary mail target node". Is it cluster alias or OA$NODE
    depending on if it's a heterogenous or a homogenous cluster?
    
    Just trying to get the picture,
    Hans

    OA$PRIMARY_NODE can have three values:
    
    IF you have a remote message router THEN
       OA$PRIMARY_NODE = OA$MTI_MR_NODE
    ELSE
       IF SYS$CLUSTER_NODE is defined
          OA$PRIMARY_NODE = SYS$CLUSTER_NODE (alias nodename)
       ELSE
          OA$PRIMARY_NODE = SYS$NODE
       ENDIF
    ENDIF
    
    Sjaak.

    .-1 describes it well. For the actual value you'll get on your site,
    look at the code in {SYS$MANAGER|SYS$STARTUP}:A1V{23|24|30}START.COM
    and see what it'll do for you.
    
    Historically, OA$PRIMARY_NODE was invented before Cluster Aliases were
    available (and supported by Message Router), in order that we could
    reliably (??) work out a node to use for the return address of a
    message. It used to be (V2.2?) set to the installing node, so that all
    returned mail would go to your most likely to be up node. Nowadays it's
    mostly the same as the cluster alias.
    
    There is a view that we could drop it and always use the cluster alias,
    but we haven't been that brave yet...
    
    Graham

    But in this case it seems they should not use cluster alias
    so therefore OA$PRIMARY_NODE can be redefined to SYS$NODE?
    
    Hans

Re .6 onwards

The confusion over the use of OA$PRIMARY_NODE is known and, to give the standard
quotation "may be fixed in a possible future release".

Scott

    Re: Hans,

    To force that the nodename instead of the aliasname is being used in
    NETWORK I *think* it's the best to do. Add a line to
    A1V{23|24|30}_SITE_START.COM that reads:

    $ nodename = "''F$TRNLNM( "SYS$NODE" )'" - "::"
    $ DEFINE /EXEC /NOLOG /SYSTEM OA$PRIMARY_NODE "''nodename'"

    Sjaak.

    We haven another problem with the running of the UA. 
    My error message is:
    
    error while opening 'filename'
    ?CANNOT OPEN FILE
    
    Is there anything I should set up before I can run the UA. I've tried
    it between two systems and it went wrong both ways.
    
    Any suggestions?
    
    Thanks in advance,
    Arjen