T.R | Title | User | Personal Name | Date | Lines |
---|
26.1 | yes | DEBUG::GALLO | Paul Gallo (CEA/SPS Support) | Thu Jan 26 1989 15:16 | 8 |
| Yes!
1) Stop the process DECW$SERVER_0
2) invoke DECW$STARTUP.COM
3) logout
Windowing will then appear!
|
26.2 | Use RESTART | DECWIN::FISHER | Burns Fisher 381-1466, ZKO3-4/W23 | Fri Jan 27 1989 12:59 | 12 |
| I think you skipped a couple steps in .1 like logging in via set host from
another terminal/workstation. As soon as you stop the server, your terminal
emulator windows go away.
A much easier thing to do would be to type @SYS$MANAGER:DECW$STARTUP RESTART
in the terminal emulator.
BTW, stopping DECnet from a window should not stop DECwindows, since you
should be running the terminal emulator over a local (non-DECNET) connection.
Burns
|
26.3 | Any chance for getting a link-error tolerant DECwindows? | IO::MCCARTNEY | James T. McCartney III - DTN 381-2244 ZK02-2/N24 | Fri Jan 27 1989 17:17 | 20 |
|
Has anyone given thought to how one might make DECwindows more tolerant of link
level failures? It seems that the Xlib and transport layers could have logic
that would allow it to reestablish context if the link dropped. What's the
current thinking on this? It doen't seem reasonable to force each application
to build the redundancy when it could be embedded in the transport.
As an example, I know of one application designed to keep track of mission
critical telemetry data. The hardware is configured such that every display
station is dual-rail ethernet, connecting to redundant hosts. The current
design calls for updates to occur on a near-realtime basis and that link
failures (due to cable or host failure) be detected and failover to redundant
systems to occur w/o interruption of the display. The time allowed for failover
is much less than the time allowed for the application to reestablish context
from an empty screen. Is there any possibility that such robust fail-safe
mechanism be built into V2.0?
James
|
26.4 | DECwindows networking leaves something to be desired | HACKIN::MACKIN | Men for Parthenogenesis | Mon Jan 30 1989 10:57 | 6 |
| Agreed. At the very least it would be nice if DECwindows used the NCP timeout
parameters in determining when to lose a link. We have an LaVC here and a
lot of the applications run on remote machines. If the DECnet circuit bounces,
even for a second, all of the applications lose their links and have to be
restarted. Even if the NCP timeout is 45 seconds or something reasonable.
|
26.5 | Not likely anytime soon | STAR::BRANDENBERG | Intelligence - just a good party trick? | Mon Jan 30 1989 11:34 | 15 |
|
In the case where a link truly disappears, which context do you
maintain? The one the server had created up to the last executed X
function or the one the client application *thought* it had based on
what it had called in Xlib/Toolkit routines (the difference being what
was buffered in the server, the client, or in the network)? And how to
recover? Will the client assist in checkpointing based on sequence
numbers or is Xlib expected to recover everything?
At present, X is merely a way of distributing graphical operations, not
creating fault-tolerant distributed systems. That is an area for RPC
services.
monty
|
26.6 | There may be a way to do this already... | DECWIN::ROSENBLUM | | Thu Feb 02 1989 11:08 | 13 |
| X has a feature that allows an applications resources and windows to stay
arround after the client dies. You could use this mode to leave the server
context intact and have the application just switch over to using a
second connection. Since all gc's and windows and such are still there
and in the exact state they were when the connection died the application
just needs to change the value of the display variable and his application
should continue to run un-interrupted.
The one thing that you must be carefull about is when the application realy
exits it must either clean up after itself, or shut off the resource save mode.
Mike
|
26.7 | Needs work | DECWIN::FISHER | Burns Fisher 381-1466, ZKO3-4/W23 | Thu Feb 02 1989 13:45 | 20 |
| re .6: Ouch! What Mike says is an interesting idea, but you have to be real
careful to avoid having your resources stay in the server forever (at least
until you log out). For example, you'd better have a kernal-mode exit handler
to make SURE you don't get shut down without cleaning up resources. Note,
however, that you can't call Xlib from kernal mode. Sounds like trouble.
In addition, you have to keep track of whether you have restarted or not.
If you have not restarted, just switching modes from RetainPermanent to
DestroyAll will do before you exit. However, if you have restarted the
connection, that means that you are no longer the owner of the resources.
You need to do a KillClient on them to make sure they go away.
Finally, I'm not sure how XLib and the tool kit will deal with this. It may
be that Xlib will destroy its various structures which are associated with
the resources that are being retained and therefore, you will never be able
to reference them again. Not to mention the toolkit.
An interesting idea, though...
Burns
|
26.8 | When/Where is X$X0 created? | OIWS20::BRYSON | | Fri Feb 24 1989 01:37 | 6 |
| Which part of the server actually creates the X$X0 object?
SERVER_MAIN? DIX?
David
|
26.9 | | JAMMER::JACK | Marty Jack | Fri Feb 24 1989 09:10 | 4 |
| XPORT_DECNET.B32, which is part of DECW$TRANSPORT_DECNET.EXE, which is
merged image activated to the server during startup. It's in facility
DIXXPORT.
|
26.10 | Many thanks! | OIWS20::BRYSON | | Fri Feb 24 1989 09:26 | 2 |
|
|
26.11 | I don't get a X$X0 object at all | DSTEG::LUND | | Mon Mar 13 1989 14:11 | 10 |
| re: .2
I have tried @SYS$MANAGER:DECW$STARTUP RESTART, and it restarts
the server process, but I still don't get any X$X0 object. I'm
running SDC DECwindows 5.1 and VMS 5.1.
Any suggestions?
-- Stan
|
26.12 | Some tests | OIWS20::BRYSON | | Mon Mar 13 1989 14:45 | 21 |
| 1. Are you using a logical DECW$IGNORE_DECNET?
2. Do a SHOW LOGICAL DECW$SERVER_TRANSPORTS? You should see LOCAL
and DECNET. The logical is located in the DECW$LOGICAL_NAMES
table.
3. When you bootup, does a message appear that it is waiting for DECNET
and then does it die?
4. The X$X0 object is used by the server and must exist on the machine
which is running the server. A non-workstation which runs a client
does not need the X$X0 object to communicate with the remote display.
Silly question but are you looking for X$X0 on a non-workstation?
(I just had to check)
5. Do an LIST under INSTALL and see if DECW$TRANSPORT_DECNET is
installed "Open Hdr Shar Lnkbl".
David
|
26.13 | RE: .12 | DSTEG::LUND | | Mon Mar 13 1989 16:31 | 35 |
|
RE: .12
1. DECW$IGNORE_DECNET - No, I don't have that defined
2. DECW$SERVER_TRANSPORTS - It wasn't defined. I defined it, then
I restarted the server with @DECW$STARTUP RESTART, and
I still don't have any X$X0 object.
3. When I boot, I get the message that DECwindows is waiting for
DECnet to start. This is because I use a SUBMIT for
STARTNET, and an "@" for DECW$STARTUP. Even when I SUBMIT
them in the same submit command, I don't get a X$X0
object.
4. I'm looking for the X$X0 object on the workstation that I want
to display the window on (a VSII).
5. Under INSTALL, a LIST command shows me DECW$TRANSPORT_DECNET
is installed OPEN HDR SHAR LNKBL
Does it make a difference what fashion I use to start up DECnet
and DECwindows?
$SUBMIT/NOLOG SYS$MANAGER:STARTNET.COM, -
SYS$MANAGER:DECW$STARTUP.COM
--or--
$SUBMIT/NOLOG SYS$MANAGER:STARTNET.COM
$@SYS$MANAGER:DECW$STARTUP.COM
--or--
$@SYS$MANAGER:STARTNET.COM
$@SYS$MANAGER:DECW$STARTUP.COM
Thanks
Stan
|
26.14 | Strange.. | OIWS20::BRYSON | | Mon Mar 13 1989 18:24 | 15 |
| No. The command procedure handles it. I submit STARTNET.COM and
execute DECW$STARTUP.COM.
Was this an upgrade to V5.1 from a FT or a straight installed from
v5.0-2?
Do you have more than one version of DECW$TRANSPORT_DECNET.EXE? Is
the creation date 17-DEC-1988 15:31:51.68?
Is the checksum for the entire image %X'87EC2D80?
David
|
26.15 | RE: .14 | KAMERA::LUND | 35mm.... | Tue Mar 14 1989 09:13 | 114 |
|
RE: .14 by OIWS20::BRYSON
This was an upgrade from VMS V5.0-2 to 5.1, no field test software was
involved.
I only have one version of DECW$TRANSPORT_DECNET.EXE (shown below). I
don't know how to determine the checksum of it.
Following the <FF> are what DECnet objects are defined, what DECwindows
logicals are there, and a DIR/FULL of DECW$TRANSPORT_DECNET.EXE.
I've check the known object list on several other nodes in my group,
and most have the X$X0 object there. There are others besides mine,
however, that don't have it.
So far, the help has been great. Keep it coming!
Thanks
Stan
$ ncp sho known obj
Known Object Volatile Summary as of 14-MAR-1989 08:57:05
Object Number File/PID User Id Password
$MOM 0
$NICONFIG 0
DFS$COM_ACP 0 00000050
TASK 0 NONE
FAL 17 FAL.EXE
HLD 18
NML 19 NML.EXE
REMACP 23 0000004D
MIRROR 25
EVL 26 0000004B
MAIL 27 MAIL_SERVER.EXE
PHONE 29 PHONE.EXE
CTERM 42 0000004D
VPM 51 VPM.EXE
DTR 63
DQS 66 DQS$SERVER.EXE
DECW$CLIENT 222
$ sho log decw$*
(LNM$PROCESS_TABLE)
(LNM$JOB_80336ED0)
(LNM$GROUP_000001)
(LNM$SYSTEM_TABLE)
"DECW$DECTERM_MAILBOX_KAMERA::0.0" = "_MBA570:"
"DECW$SERVER_MBX_0000FFFE_0" = "MBA250:"
(DECW$LOGICAL_NAMES)
"DECW$BOOK" = "SYS$COMMON:[DECW$BOOK]"
"DECW$EXAMPLES" = "SYS$COMMON:[SYSHLP.EXAMPLES.DECW]"
"DECW$INCLUDE" = "SYS$COMMON:[DECW$INCLUDE]"
"DECW$KEYMAP" = "SYS$COMMON:[SYS$KEYMAP.DECW.USER]"
= "SYS$COMMON:[SYS$KEYMAP.DECW.SYSTEM]"
"DECW$SERVER_TRANSPORTS" [super] = "DECNET"
= "LOCAL"
"DECW$SERVER_TRANSPORTS" [exec] = "DECNET"
= "LOCAL"
"DECW$STARVUEECOM" = "SYS$MANAGER:DECW$STARTVUE.COM"
"DECW$SYSTEM_DEFAULTS" = "SYS$COMMON:[DECW$DEFAULTS.USER]"
= "SYS$COMMON:[DECW$DEFAULTS.SYSTEM]"
= "SYS$LIBRARY:"
"DECW$USER_DEFAULTS" = "SYS$LOGIN:"
"DECW$WINMGREXE" = "SYS$SYSTEM:DECW$WINMGR.EXE"
"DECW$XLIBERRDB" = "SYS$MESSAGE:DECW$XLIBERRDB.DAT"
Directory DUA0:[SYS0.SYSCOMMON.SYSLIB]
DECW$TRANSPORT_DECNET.EXE;3 File ID: (1147,28,0)
Size: 17/18 Owner: [SYSTEM]
Created: 17-DEC-1988 15:31:51.68
Revised: 13-MAR-1989 15:42:16.41 (4)
Expires: <None specified>
Backup: <No backup recorded>
File organization: Sequential
File attributes: Allocation: 18, Extend: 0, Global buffer count: 0
No version limit, Contiguous best try
Record format: Fixed length 512 byte records
Record attributes: None
RMS attributes: None
Journaling enabled: None
File protection: System:RWED, Owner:RWED, Group:RWED, World:RE
Access Cntrl List: None
Total of 1 file, 17/18 blocks.
|
26.16 | DECW$SERVER_TRANSPORTS not defined... | OIWS20::BRYSON | | Tue Mar 14 1989 13:05 | 44 |
| RE: .13
> 2.DECW$SERVER_TRANSPORT - It wasn't defined. ...
That shouldn't happen. This logical is defined in
DECW$STARTSERVER.COM and there are no conditionals that should cause it
to skip this definition.
$ decw$define decw$server_transports decnet,local
Strange.
Have you made any changes to the DECwindows environment (com files)?
Look at DECW$SERVER_0_ERROR.LOG for any error information when it tries
to load the DECnet transport.
When the server come up, does everything work OK locally?
May be GBLSECTIONS or GBLPAGES are being depleted. Look in INSTALL.
This is probably not the problem.
It bugs me that DECW$SERVER_TRANSPORTS was not defined. Try this.
1. Log into the system from another terminal
2. Stop the current Server process
3. DEFINE DECW$SUBPROCESS_IGNORE "TRUE"
4. Make sure that DECnet is up and running
5. SET VERIFY
6. @DECW$STARTUP
This will give you output from each step of the startup process.
Keep a close eye out for error messages and when DECW$SERVER_TRANSPORTS
should be defined in DECW$STARTSERVER.COM
David
|
26.17 | | ATPS::MALLORY | VAX SPM Development - Home of DECcp | Wed Mar 15 1989 08:20 | 8 |
| Look and see if you're running out of global sections... I
just had an experience where a NEW installation of DECwindows (even
after an autogen with WINDOW_SYSTEM=1 caused there to not be enough
global sections (probably due to other layered products) and the
DECwindows startup quietly died.
|
26.18 | the missing object is no longer missing | LAS057::LUND | 35mm.... | Wed Mar 15 1989 13:57 | 15 |
|
RE: X$X0 object never there
The problem isn't there anymore. Out of desperation, I
re-installed DECwindows, and the problem disappeared. I couldn't
afford to take any more time flailing at it.
Thanks to everyone for the suggestions. Without this (and other)
conferences, I don't know where I would turn.
-- Stan
|
26.19 | Yes, I have no X$X0 . . . | NANOVX::ROBERTS | Keith Roberts LKG2-2/N1 | Wed Mar 15 1989 15:39 | 16 |
| I have been having the same problem -- no X$X0 object defined. I first
noticed the problem when running 5.1-t2 and decw bl13. I had been
using decwindows across the network, with no problems...then all of the
sudden the client reported "Toolkit error: Can't Open Display". The
security stuff under the session manager was correct.
The only thing I could think of that changed was disabling TASK 0
(corrective action for the virus).
I hoped that upgrading to sdc VMS & DECwindows would fix the problems -
but it hasn't.
Could the TASK 0 have anything to do with it?
Keith
|
26.20 | Nope. | STAR::BRANDENBERG | Intelligence - just a good party trick? | Thu Mar 16 1989 10:32 | 6 |
|
If you're running sdc decwindows, the task object will not affect
anything.
m
|
26.21 | There has got to be a simple answer . . . | BOEHM::D_ROBERTS | Keith Roberts LKG2-2/N1 | Fri Mar 24 1989 07:35 | 15 |
| So then - no other hope of finding out why my task X$X0 is dead?
I see from an earlier reply, that someone just 're-installed'
DECwindows (again) and that fixed the No X$X0 Object problem. Our's
is an LAVC, and reinstalling DECwindows would take down six
workstations for the better part of a day. I guess I can't see what
installing the same kit would do. We upgraded another standalone
workstation from the *same* vms & decw kits - and they have the network
object *and* have TASK Object disabled too!
Anyone have any ideas ?
Thanks,
Keith
|
26.22 | Fixed ! | TOOK::D_ROBERTS | Keith Roberts LKG2-2/N1 | Wed Mar 29 1989 16:36 | 5 |
| The bug has been found ... read 19.26 for the answer !!!
Keith
|