[Info-vax] grey screen of death

[Phillip Helbig undress to reply]

In article <5561e787$1 at news.langstoeger.at>, peter at langstoeger.at (Peter
'EPLAN' LANGSTOeGER) writes: 

> *) You have a dying X11 server on your Alpha

It's not working; whether it is "dying" I don't know.

> *) You haven't found DECW$SERVER*.DMP files or there are none

Right.

> *) You haven't found related accounting records or there are none

The DECW$SERVER_0 or whatever process creates the DTLOGIN process.  This 
lasts just a couple of seconds.  It seems to produce the "request for 
dump" or whatever in the auditing file.  The DECW$SERVER_0 process does 
this every few seconds for a few minutes then stops.  During the loop, 
the grey screen of death changes to the black background with the 
hourglass, which is normally what appears immediately before the CDE 
login box comes up.

> *) Without an X11 server, you have no CDE login/greet process (of course
> and no DECW$TE* terminal emulator process of DECterms of logged in user
> and also not the interactive process of this logged in user

Right.

> because you are not logged in via the CDE login screen at all) 

Right.

> *) You are running OpenVMS V8.4 with all ECOs (or are in process to install
> the latest VMS ECOs)

No.  I have UPDATE5.  Normally, when there are problems, the first 
question is whether all patches are up to date.  Obviously, as a 
hobbyist I don't have access to newer patches (UPDATE10 is out now) and 
don't want to violate any terms of the hobbyist license.

Of course, if someone responsible says that it is OK for me to install a 
newer patch and see what happens, then I'm willing to do that, but I 
would prefer to stick with the official hobbyist kit.

Maybe someone with access to the 8.4 patch release notes can see if 
there is any thing related to looping processes involving CDE and/or 
DECwindows and if so let me know which patch it is in.  Maybe a newer 
UPDATE can then be made available for hobbyists.  Someone in this thread 
posted something about a looping DECW process on 8.4, which is what I 
have.

> *) You are running TCPIP V5.7 ECO5 (or are in process to install ECO5)

Yes.

> *) You are running DECwindows MOTIF V1.2-7 w/o ECOs (because there are none)

I don't recall the version, but it is that on the 8.4 distribution "CD"
(I actually downloaded the .ISO, copied it to a disk, and booted from
that for the upgrade).  It is a newer version than I had before the 
upgrade.

> *) You are not running DECnet (neither Ph4 nor DECnet-Plus/DECnet-OSI/Ph5)

No DECnet.

> *) You have working DECwindows client, as you could prove with an X11 server
> on a remote system (like another of your OpenVMS systems with a working
> X11 server or like a PC with a X11 server like eXcursion or eXceed or ...)
> but you have not done so

I did do RUN SYS$SYSTEM:DECW$CLOCK on the 8.4 system with the display 
directed to a 7.3-2 system.  This works.

> My suggestion is:
> *) Get yourself a PC with a working X11 server (and login via SSH/Putty
> and X11 forwarding enabled and then CREATE/TERMINAL/DETACHED to get a DTTERM
> or use the XDM functionality of the X11 server to login via DTGREET/DTLOGIN)
> and prove that your problem is only your OpenVMS DECW$SERVER_0 process

I'll check this in more detail.  I do remember at one time bringing up 
not just a DECTerm window but the whole CDE login box on a remote 
display.

> *) Complete your upgrade activities (including DECnet installation ;-)

I think everyone agrees that DECnet is not relevant here.  I do plan to 
get it running, hopefully this year, but I think it's not a good idea to 
change too many things at once.

> 	check SYS$STARTUP:DECW$SERVER_0_ERROR.LOG

Nothing there.  Essentially the same as with 7.3-2.

> 	check ACCOUNTING for data of DECW$SERVER_0 process (username SYSTEM)

Have done so and posted the -F- message.

> -) increase debugging info
> 	by adding following lines to SYS$STARTUP:DECW$PRIVATE_SERVER_SETUP.COM
> 	$ decw$server_connect_log == "T"
> 	$ decw$server_dump == "T"

I'll give that a try.

> 	and by using AVAILMAN (and watch the process quotas)

I don't think I have that installed.  Even if I did, I have no 
experience with it.

> -) increase your (DECW$SERVER_0 process and system) quotas
> 	by adding 2 lines to SYS$STARTUP:DECW$PRIVATE_SERVER_SETUP.COM
> 	$ DEFINE/NOLOG DECW$SERVER_PAGE_FILE   <at least 370000>
> 	$ DEFINE/NOLOG DECW$SERVER_WSQUOTA     <at least 12288>

Will do if they aren't this high already.

> 	and by considering the following lines for your MODPARAMS.DAT
> 	MIN_LNMSHASHTBL=512
> 	MIN_MAXPROCESSCNT=128
> 	MIN_PQL_DBYTLM=32768
> 	MIN_RESHASHTBL=512
> 	MIN_TTY_ALTYPAHD=2064
> 	MIN_TTY_ALTALARM=1600
> 	MIN_TTY_TYPAHDSZ=2064

Ditto.

> -) reduce your X11 server to LOCAL transport only
>         by changing the line in SYS$STARTUP:DECW$PRIVATE_SERVER_SETUP.COM
> 	$ decw$server_transports == "LOCAL"
> 	and see if it makes a difference

OK.

I'll probably try this all later tonight or tomorrow morning.  If none 
of this works, I probably won't have a chance to get it solved by 
tomorrow night, so will have to wait until the end of July.

Again, someone did mention a looping DECwindows server in 8.4.  This 
seems like a typical thing a patch would correct.  I suspect that most 
8.4 early adopters were not running DECwindows, so it might have been 
fixed after UPDATE5.  If someone can find something relevant in the 
release notes for newer patches, that would be the obvious avenue to 
explore, and maybe the hobbyists can get a newer update made available.