Conference iosg::all-in-1_v30

    
    Sunil,
    
    are you sure it isn't the FCS that has these files locked - i'm
    guessing since hundreds of your other notes last week are about the FCS
    
    Regards,
    
    Andrew.D.wicks

    Hi Andrew,        
    
    Does the FCS access PENDING.DAT and ATTENDEE_OVERFLOW.DAT ?  I have
    some doubts about ATTENDEE_OVERFLOW.DAT, maybe Bob (Also known as Mr.
    FCS) can confirm this one way or the other !!!
    
    I made a slight mistake when I asked the customer to do a $ show
    device/files oa$data, I should have asked the customer to use SYSMAN.
    
    mc sysman
    set environment/cluster
    do show device/files oa$data
    
    This would give me a better picture of what is happening on the other
    nodes in the cluster.
    
    Apart from the "does FSC ..." question I have asked I have another one.
    When ALL-IN-1 shutsdown it's shutdown on all nodes in the cluster and
    users who are logged in are gently returned to the DCL prompt.  Would
    any condition prevent this from happening ?  I'll hopeful be able to
    get some more details tomorrow the customer will run the RSF tonight.
    
    Regards,
    
    Sunil
    
    PS - Andrew "U.S.A 2 England 0 - I was there!" you have forgotten the
    ":-)".  Anyway they play soccer not football.  Andrew the Welsh dragon 
    smiles again !!!!

Hello all,

>Does the FCS access PENDING.DAT and ATTENDEE_OVERFLOW.DAT ?  I have

Must have been a user still logged in. Easy to reproduce by pressing
Hold Screen in an ALL-IN-1 session and running RSF. AST will be delivered,
so no "Some users could not be notified of the shutdown" message
in the log file. But the image wil not exit and the converts on 
both files fail.

Regards,

Jan

    Hi Jan,
    
    Yes I know about the old "Hold Screen" and the Stars article stuff the
    problem was discovered at a customer site DownUnder.  But the customer
    assures me that's not the case, I don't trust some of our customers
    hence the show device/files.  Anything else that would cause the
    problem we'll find out soon ....
    
    This problem has been happening since May and was only reported last
    week.  Now could they have one user who always has his hold key down or
    disconnects from the server without logging out ?
    
    I hope this problem is fixed in a PFR.
    
    Regards,
    
    Sunil

    
    THe FCS does access PENDING.DAT, but not ATTENDEE_OVERFLOW.DAT (I
    don't even know what that one is :-) ).
    
    One thing that could keep users from being gently kicked out during
    a shutdown, is when the FCS shutsdown, it will let users finish the
    task they are doing. So, if for example, that task was to copy a remote
    drawer to another remote drawer and there were millions (or some big
    number) of documents going on, and the network was particularly slow
    that day, it would appear that the user missed the "get out" message.
    When in fact, he eventually would exit, it just may take awhile.
    
    --Bob
    

    Hi Sunil,
    
    Another possibility: there's a known problem that the Script Symbiont
    doesn't stop for a system shutdown. Does your customer use it?
    
    Cheers
    
    Stuart

    Hi Stuart,
         
    Thanks for your suggestion the one possibility that we missed was .....
    
    Well one of their users had defined a symbol
    
    a1:==allin1/reenter/override/user="xyz"
    
    It was this one user who had the symbol setup with the /override that
    caused the problem.  It was obvious of course 8^).  Need I say the
    system mangler had a few things to say to the user.
    
    Regards,
    
    Sunil

    I assume the user in question isn't an an administrator, nor privileged
    anymore??  :-)
    
    Graham

    TRAKTEREN !!!!!!
    
    I don't know how hard it is to notice the ALL-IN-1 manager
    about users beeing logged in by the shutdown command like
    it does when someone is in a spawn-mode.
    
    Regards,
    
    Jan

    Hi Stuart and Jan,
    
    Well the customer was a bit embrassed and didn't say anything apart
    from please close the call :-).
    
    One question I do have is concerning what Jan has mentioned about if a
    user has spawned than the shutdown procedure will not logout the user.
    Does it kill off the parent process and leave the spawned process
    running ?  What's the reason behind that and any work arounds in place
    to overcome the problem ?
    
    Regards,
    
    Sunil

    Hi Sunil,
    
>   What's the reason behind that and any work arounds in place
>   to overcome the problem ?
    
    It's do with ASTs not being delivered whilst the user is spawned. There
    are some home-grown procedures around to kill off these processes. The
    one we use on IOSG is a command procedure that finds out who has
    MEMRES.FLB open, and does a stop/id on them.
    
    Cheers
    
    Stuart

    The shutdown procedure uses the lock manager (as it works across nodes
    in a cluster; this is also one reason why the process needs SYSLCK
    priv). It uses a feature called "blocking ASTs" which let one process
    tell all other processes which are holding a given lock that it's
    required. This happens by delivering an AST to each process.
    
    Unfortunately, for reasons best known to VMS-land, when a process is at
    DCL (because of a control-Y, or SPAWN or ATTACH to another process),
    user-mode AST delivery is inhibited. So the shutdown "message" doesn't
    get delivered to such processes. The process doing the shutdown wits
    for 30 seconds for the deliveries to take place and then tells the
    Manager that it couldn't get through to some users.
    
    The only way out of this is to use a higher access mode than user-mode.
    An attempt was made to put the entire shutdown mechanism into
    Supervisor mode, but this turns out to be amazingly difficult for all
    sorts of other reasons and had to be abandoned. We did change control-Y
    so it effectively did an <EXIT to get rid of the most common case.
    
    Note that this is also the reason for the odd "hangs" you sometimes get
    (not as often now) with DDS, as it uses the same mechanism to do
    SUSPENDs. This doesn't cause problems with SPAWN though as we release DDS
    before the SPAWN and re-connect to it afterwards.
    
    Here endeth the history lesson!
    Dave.

    PS - I like the idea of detecting users who are running /OVER. This
    ought to be possible; a "wish"?
    
    Dave.