Conference lassie::ucx

5280.0. "Communication proxy table doesn't work" by RULLE::LUNDEBORG () Wed Feb 26 1997 08:17

    
    A customer has problem with communication proxies. When he starts UCX
    it looks like all proxies are read in, but if you do a "sho proxy" in
    UCX you can see that some proxies aren't loaded. If you at this point 
    do a "add proxy" for these proxies which wasn't loaded you can see that
    some of them will be loaded this time and in the same time some other 
    proxies will become unloaded (not cached).
    
    The start of UCX looks like this:
    
    
    %UCX-I-LOADSERV, Loading UCX Server proxy information
    %UCX-I-SERVLOADED, UCX Aux. Server loaded with 3278 proxy records
    -UCX-I-SERVSKIP,  Skipped 0 communication proxy records
    -UCX-I-SERVTOTAL,  Total of 3278 proxy records read
    
    if you now for proxies you will find some of them missing the "C" in the
    type-field and also if you try that user the proxy doesn't work.
    
    In this example they had 202 entries which not was loaded, but if you
    restart UCX one more time it can be less which not load.
    
    They run in a cluster which looks like follows:
    
    Config:
    * 3 CI VAX-cpu's named BLGV04,BLGV06 and BLGV07
    * every cpu has it's own sysuaf.dat and ucx$proxy.dat
    * BLGV04 has its own system disk
    * BLGV06 and BLGV07 share systemdisk
    * all machines runs VMS V6.2 and UCX V4.1 ECO-2
    
    They have seen the problem on both BLGV04 and BLGV07 but so far
    has the problem been at only one node at a time.
    The problem started on node BLGV07, and when we after a lot of 
    trying on this node (by making proxyentries fewer), we had all 
    the proxy-entries installed, this felt great. The bad thing was
    that BLGV04 was down at this time and when it came up we had the
    old problem here, that some proxies weren't loaded.
    
    Outsatnding questions:
    
    * Are we running an unsupported way? 
    * Is it a maximum value of remote proxies? (I've put in 5000 proxies in
      my own machine and it works ok)
    * If we are running out of some kind of resource, where to look?
    * Is it UCX$INETACP to blame or ......
    
    
    
    
    		/Dan Lundeborg  TSC Sweden

    
    Naturally I ment the "D" bit and not the "C" bit as I said in .-1
    Sorry for that  
    
    	/Dan Lundeborg

    
    One more thing:  They are running Pathworks on all nodes via TCP/IP and
    		     they have adjusted INTSTKPAGES up to 16 (from def 6)
                     
    
    
    			/Dan Lundeborg