Re: crs farm

From: Christian Holm Christensen (cholm@hehi03.nbi.dk)
Date: Mon May 06 2002 - 08:23:12 EDT

  • Next message: Djamel Ouerdane: "raw data output"

    Hi Flemming, 
    
    On Thu, 2 May 2002 18:37:48 -0400
    "Flemming Videbaek" <videbaek@sgs1.hirg.bnl.gov> wrote
    concerning "crs farm":
    > For everyone information.
    > 
    > The problem observed by Djam and Pawel was in the end traced to the
    > fact that a subset of nodes 0008-00012 are still running RH6.2 thus
    > /opt/brahms pointing to different physical directories.  
    
    Which ofcourse would break things badly. 
    
    > I may have heard this but not registered.The reason is these machine
    > have both less memory, local disks and are slated for removal during
    > the summer. The action for  now is to disable these from crs until
    > put into another queue. 
    
    Which means, however, that you should _not_ use the drop-queue
    feature, as it can happen that your job gets pushed onto a Red Hat 6.2
    system rather than a Red Hat 7.2 system.  Heads up on that one.
    
    > The second morale is that the first conclusion reached on
    > misbehaving system ought not be the right answer after all AFS did
    > what it was supposed to do, and Brat probably build correctly. 
    
    BRAT was not updated correctly - there, that's said.  One should
    always uninstall the previous minor version, as you can really get
    into some oddities if you don't, since the minor version number isn't
    used in the soname.   
    
    Also, you could end up with a huge amount of old libraries lying
    around.  All the BRAT libraries takes up 23M in total.  
    
    Anyway, the proper people know this, as I mentioned it several times. 
    
    The kernel does not unload the shared libraries unles it really needs
    to, so if you're logged on over an upgrade, the old libraries may be
    stuck in the cache.  A sync will help you.   
    
    The best way to avoid all these troubles was if we had individual
    version numbers for each library.  However, when we made BRAT 2, I
    believed that it would easily be forgotton, which would lead to much
    more confussion.  If the sentiment is that it will not, we can chagne
    it. 
    
    > The first morale is to read in details the message that come from
    > RCF, to catch these kind of exceptions 
    
    I generally read the mails from RCF on these topics (in the form of a
    digist), but I didn't see any mail that said that rcas0008-12 are
    running Red Hat 6.2 - I just browsed the archive and found - nothing. 
    
    > cheer up
    
    ditto. 
    
    Yours, 
    
     ____ |  Christian Holm Christensen 
      |_| |	 -------------------------------------------------------------
        | |	 Address: Sankt Hansgade 23, 1. th.  Phone:  (+45) 35 35 96 91
         _|	          DK-2200 Copenhagen N       Cell:   (+45) 24 61 85 91
        _|	          Denmark                    Office: (+45) 353  25 305
     ____|	 Email:   cholm@nbi.dk               Web:    www.nbi.dk/~cholm
     | |
        
    



    This archive was generated by hypermail 2b30 : Mon May 06 2002 - 08:23:46 EDT