Conference marvin::x25psi

3749.0. "Pending AST problems on X25 PVC" by ATHINA::TSAKALOS () Thu Apr 24 1997 15:09

               
               
               
               Hello all,
               
               We are kindly for your expertise and help in solving a 
               problem involving a network process and PSI interface 
               for X.25/PVC connections.
               
               The network processes consist of the network server, 
               which is a unique process and creates (detached) 
               network agents.  Each network agent is responsible to 
               handle multiple PVCs (up to 10).
               
               The problem we encountered is for AST not completed, 
               causing the agent process to go into SSRWAIT state, 
               when we try to shut down the application.
               
               We have observed that the problem occurs after a 
               collision of a data packet sent by network agent 
               (DATA.request and a RESET.indication sent by the ATM.
               
               In this case, the AST of DATA.request is pending, 
               while the AST of RESET.response is completed, although 
               it was issued after the DATA.request.
               
               The network agent process can continue it's work even 
               after this situation, but the DATA.request AST never 
               completes.
               
               But, if someone tries to shut down the network agent 
               process, this one goes to SSRWAIT state and hangs out.  
               The only thing we can do is to REBOOT the system. To 
               be more specific, the hanging of network agent happens 
               after $QIO (IO$_DEACCESS) on the problematic channel, 
               although this system service returns SS$_NORMAL end 
               the PVC is free for reuse.
               
               We also have noticed that there are two NWAnn: devices 
               allocated. These are, one for the PVC (that was 
               deaccessed!!!) and one for network declaration channel 
               (IO$_ACPCONTROL) that didn't had the chance to be 
               deassigned, because of the hang out.
               
               Even when we trid to shutdown and restart PSI again, 
               these devices weren't deallocated.
               
               But, if we try to run another network agent process, 
               the PVC is given, associated to another NWAnn: device 
               and we can communicate normally.
               
               We need your help to overcome this problem without 
               rebooting the system.
               
               Attached there is a scheme with the above operations
               
               I will be happy to provide any additional information, 
               if you need it.
               
                Thanking you in advance
                 Best Regards
                 Thanos


                  ATMCOM                      ATM
                       |  RESET.REQUEST        |           
                       |--------\              |
                       |         \____________>|
                       |                       |
                       |   RESET.confirm      /|
                       |<--------------------/ |
   AST.DATA.request    |                       |
      pending          |>------\               |  
                       |        \    / -------<|   
                       |     data Requests     |                             
                       |         \-/---------->|
                       |          /            |
AST.RESET.responce     |<--------/ reset.indic |   
    pending            |\                      |                            
                       | \  reset.response     |    
                       |  \___________________>|
                       |                      /|     
                       |   delivery          / |                          
AST.RESET.responce     |<--------------------  |
  completion             AST.reset.responce
                           
                                  
NOTE: I Have enter the same on notes DECNET-OSI_for_vms (3944)

    
        Hi all,
        
         I forgot to tell you ,we are under VMS V6.2 and DECnet/OSI
         V6.3 ECO 6
    
        Thanks 
      Thanos

    Is this a direct connection (LAPB or LLC2) or are you going via GAP?
    
    Antonio

    
   Hi Antonnio,
    
   The configuration is LLC2 PVC to LAPB PVC connected via
   remote X25 RELAY PVC. This is for MTB bank for which you
   produced the DECnis patch NIS031.49956
    
   Argyris.

It sounds like you need to log an IPMT. What we need is:

	1. Directions for reproducing the problem
	2. A minimal program to reproduce the problem

Thanks

Antonio