🚀 go-pugleaf

This message is being cross-posted to vmsnet.sysmgt, comp.os.vms and
vmsnet.networks.misc. Please post reposnses in vmsnet.sysmgt or e-mail
the author.

People from WordPerfect Corp. are encourraged to read and respond to 
this post (see below).

It would appear that this problem I have perviously described is alot 
more widespread than I thought. I have received a number of responses 
to my query; many from sites that have experienced this problem now or 
in the past. For this  reason, I would like to return this discussion 
to the group.

I summarize the problem again along with some of the latest information
collected.
 
From time to time, we will get one or more interactive terminal sessions on 
LTAxxxx: terminals getting "lost in space". We have not been able to 
reproduce this problem in a controlled fashion. However, the affected 
sessions show ALL of the following symptoms:


- A SHOW USERS FULL will be missing the (server/port) address for the
LTAxxx: terminal
$ SH US /FU SEXSMITH
      VAX/VMS User Processes at 25-JAN-1993 16:45:48.57
    Total number of users = 1,  number of processes = 1

 Username     Node   Process Name    PID     Terminal
 SEXSMITH_M    VAXB  SEXSMITH_M    20601743  LTA6716: 

- The process will INHALE all available CPU time, effectively turning the 
system into a dog (it turns out the jobs are NOT at elivated priority; my 
memory had failed me). Doing a SHOW PROCESS on the job shows that it is 
locked in some sort of CPU loop, doing no I/O whatsoever; the job state is 
always COM.

- The application running is one writen and supported by us. We have seen 
two applications involved here, both written in VAX BASIC. One application 
uses SMG exclusively for terminal I/O; the other simply invokes the PRINT 
and INPUT verbs built into the VAX BASIC language.

- We have no reliable reports of what the physical terminal is displaying

- SHOW ERROR reveals nothing

- Happens for both direct and dialup connections on any port, any server, 
DEC and non-DEC. One site uses a Vitalink bridge (with the problem appearing 
on both sides); the other has nothing but a single Ethernet segment and a 
number of terminal servers.

- The problem tends to happen during peak hours. We cannot recall, but 
cannot rule out occurances during off-peak hours.

- A SHOW PORT LTAxxx command from LATCP shows that the port is inactive 
and does not show the server/port name for the device. Even more 
interesting is that the port type is indicated as 'Forward', with no 
target service or actual service indicated. This port started out as an 
interactive!

- Server error counters do not seem to indicate trouble. A SHOW PORT COUNT
command on the server indicates some framing errors, but not an excessive
number:
     Current Counters for Port 19 qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
    x 
    x Seconds since zeroed: 1780534 (20 14:35:34)
    x
    x Port Statistics:                      Port Errors:
    x   Local Accesses:           74          Framing:                 71
    x   Remote Accesses:           0          Parity:                   0
    x   Inactivity Logoff:         0          Overrun:                  0
    x   Password Logoff:           0

- NCP counters do no show anything out of the ordinary. A SHOW EXECUTOR 
COUNT display for one instance follows:
Node Counters as of 25-JAN-1993 16:45:35
 
Executor node = 1.11 (VAXB)
 
      >65534  Seconds since last zeroed
     1448613  Bytes received
     1448396  Bytes sent
       31035  Messages received
       31243  Messages sent
         208  Connects received
         208  Connects sent
           3  Response timeouts
           0  Received connect resource errors
          11  Maximum logical links active
           0  Aged packet loss
           0  Node unreachable packet loss
           0  Node out-of-range packet loss
           0  Oversized packet loss
           0  Packet format error
           0  Partial routing update loss
           0  Verification reject


**************************************************
In response to my query I have received the following:

From: "FLOWERS HARRY" <FLOWERS@memstvx1.memst.edu>
Subject: Re: Lost LAT sessions
To: "psmode" <psmode@jeslacs.wimsey.bc.ca>

We've had the same problems.  Typically with SMG-based applications,
but also happens with LISP.  I've got a command procedure that hunts
them down and kills them.  I'll include it at the end.

>So far, DEC has sent the latest patch kits and had us install them, but
>to no avail.

Yea, we went around with DEC on this, but they've decided it's a feature.
Evidently, if you try to do I/O to a LAT port after it's disconnected, this
happens.  They suggested checking all return statuses carefully, and avoiding
any I/O to a port once you get a disconnect error.  Evidently, SMG isn't
handling this correctly.  They've got a fix for VAX LISP to keep it from
happening there.  If you ever get them to admit that there's actually a
problem with either SMG or LAT software in this regard, please let me know.

--
Harry Flowers                           Internet: FLOWERS@MEMSTVX1.MEMST.EDU
Memphis State University                & Bitnet: FLOWERS@MEMSTVX1

**************************
From: grant%mighty.dnet@gw.wmich.edu (NORM GRANT)
To: psmode@jeslacs.wimsey.bc.ca
Subject: LAT looping?

I don't know if this relates to your problem or not, but we had a version
of WordPerfect for VMS which did this if a user hung up the phone while
in it.  It was a MAJOR nuisance.  Apparently the program attempted to
trap the hang up or forced exit condition and went crazy when the terminal
was gone.  You could have a similar problem, possibly with different
software.

-------------
Norman D. Grant                 INTERNET:  grant@gw.wmich.edu
Western Michigan University     Voice:     (616) 387-5430
University Computing Services
Kalamazoo, MI  49008

**************************
From:    SYSMGR@bigvax.alfred.edu (Jim Walker)
Message-Id: <930126125507.30a0373a@bigvax.alfred.edu>
Subject: LAT sessions getting lost
To: @jeslacs.wimsey.bc.ca, psmode@jeslacs.wimsey.bc.ca
X-Vmsmail-To: SMTP%"psmode@jeslacs.wimsey.bc.ca"

I have encountered that problem.  What happens to me is a user logs out of
the terminal server and VMS doesn't just delete the process any more.  It
aborts all $QIOs to the terminal with SS$_HANGUP or something like that.
Applications are supposed to detect the fatal error and exit, but some just
keep retrying and sucking up CPU time.  SHOW DEVICE LTAxxx: reports that
the device is offline.  Now I run a detached process that wakes up once a
minute and does $DEVICE_SCAN and $GETJPI looking for offline LTAxxx:s and
deletes processes.  The fortran program follows my .SIG.  It's overly
complicated because at the time I wrote it I thought it was going to
evolve into a general idle process killer.

Jim Walker
VAX System & Network manager, Alfred University Computer Center,
Alfred, NY 14802 USA        +1-607-871-2222,    Using VAX/VMS 5.4-3
<SYSMGR@bigvax.alfred.edu>, SYSMGR@CERAMICS.bitnet, WALKER@ALFREDU.bitnet

**************************************************


Any insight from WPCorp would be appreciated. Also, If somebody could 
check the fiche to see if there is any common element to the $QIOs that 
are triggered by calling input routines from SMG, VAX BASIC and LISP.


-- Peter
**************************************************************************
*   Peter Smode                 E-mail:    psmode@jeslacs.wimsey.bc.ca   *
*   JES Library Automation      Voice:     (604)939-6775                 *
*   Coquitlam, BC, CANADA       Fax:       (604)939-9970                 *
**************************************************************************