Conference csc32::consolemanager

428.0. "RWMBX for PCM resulting in 'hung'" by UTROP1::RIJSBERGEN_M (Marcel Rijsbergen @uto) Tue Oct 11 1994 15:05

    
    Warning for documentation and possible SYSGEN changes.
    
    A customer reported problems with PCM (OpenVMS V6.0, PCM V1.5, NO ECO
    yet). Didn't work at all, hung !!!
    When I was at the customer site no problems did occur for a while but
    at last yup, total PCM hung, no response at all.
    Did a '$ SHOW SYSTEM' and both 'Console Notify' and 'Console Ctrl 01'
    were in RWMBX. Couldn't work at all. Stopped, started PCM and it
    worked.
    
    After some time the same problem occurred. Now we started checking and
    walking around before we wanted to stop PCM. One terminal was connected
    to PCM via CONSOLE MONITOR, so the first system was automatically
    selected in the active window part. He put is terminal in 'lock' on the
    DECserver prompt. As soon as we restarted this window a couple of
    messages passed by in the window and PCM was working correctly again,
    RWMBX was gone. So the 'lock'-ed screen got the mailbox filled to its
    maximum resulting in the 'hung' state.
    
    There should be a warning in the documentation about this and maybe
    playing around with SYSGEN parameters (TTY_ALTYPEAHD etc.) could make
    some buffers a bit bigger so PCM does not stop that fast.
    I could not find references to the warning or the parameters.
    
    Any comments !
    
    Regards, Marcel.

Marcel,

  We will update the documentation for the next release.

  The same applies to the WATCH interface (and this is
  documented) if your output device is slow or runs out
  of paper!!

  Thanks for letting us know.

Cheers,
Phil

    >There should be a warning in the documentation about this and maybe
    >playing around with SYSGEN parameters (TTY_ALTYPEAHD etc.) could
    >make
    
    TTY_ALTYPAHD would only band-aid the problem. If somebody walks away
    from a busy system and locks the terminal it's still going to hang
    at some point as TTY_ALTYPAHD has a max value of 32K. 
    
    What needs to be done is soemthing similar to VCS. If a user
    interface can't keep up with the flow of data or is locked then 
    just move past them. If you've ever seen the VCS-W-CTYLOSS error then 
    it indicated that your interface had not kept up. No data was actually 
    lost. It was placed in the log file and the user could do a review
    and see it. This method also prevents a managed systems console from
    being XOFF'd for so long that it actually crashed or hung. 
    
    Regs,
      Dan