unable to start LPD CLIENT

[Jan van den Ende]

Hi,

yesterday we had a temporary (planned) power down on one location.
After the systems came back up, we could not start LPD queues.
The message during TCPIP startup was
%TCPIP-E-STARTFAIL, failed to start LPD_CLIENT
%TCPIP-E-NOSERVREC cannot find LPD_CLIENT service database record.

It now came to the surface that this functionality was activated some two years ago on only one node, by someone who has since long left.

I have NOT been able to find how to get this corrected !! The service manual is not much help either. :-(

VMS 7.3-2 -- ICPIP services 5.4 ECO 4.

I am sure that it will be too trivial to enter into a manual, but we are at a loss right now.

Thanks in advance!

Proost.

Have one on me.

jpe

[Volker Halle]

Jan,

the LPD_CLIENT service seems to come from TCPIP V5.1. Have a look at SYS$MANAGER:TCPIP$V51_CONVERSION.COM - it talks about merging LPD and LPD_CLIENT.

I can't find a LPD_CLIENT service entry after configuring LPD on TCPIP V5.6.

$ TCPIP SHOW CONF ENA SERVICE *

Volker.

[Wim Van den Wyngaert]

Also try Volkers command with /common to verify the enabled services. Some wisenose could have used it (I did but IMO it's not correctly implemented).

Wim

[EdgarZamora]

WAG...

try (may have to define tcpip commands first):

$ tcpip$lpd_utilities -f START_CLIENT_QUEUES

[Jan van den Ende]

Volker, Wim,
found that already. Conversion was done in 2003 (.FLG files for every node still there to prove it).

Edgar,
will try that tomorrow. Points assignment awaits the result.

Meanwhile, I have set one node to no longer accept new connections, so tomorrow I can do a full TCPIP shutdown & startup, and if that does not help a reboot.
If either fixes the issue, I will initiate a rolling down-up cycle.

Meanwhile, if anybody has more ideas (or, perhaps, documentation on this issue?), I am more than interested!

Proost.

Have one on me.

jpe

[Walter Miller_1]

On the node that LPD was ORIGINALLY activated try.

TCPIP SHOW SERVICE/PERM/FULL LPD

In a cluster only one node has to be configured but that node does have to start the service. Look for logicals that should be defined.

SHOW LO *LPD*

Especially look for the printcap file logical.

[Jan van den Ende]

Edgar,

alas.

Walter,

all _LOOKS_ correct, and not different from another cluster that is functioning OK.

The real "trouble" manifests itself when trying to start an LPD que:

Message number 07648312
after $ SET MESSA SYS$MESSAGE:TCPIP$MSG that is:
%TCPIP-E-FACNOTSTART, attempt made to run a TCPIP image before starting the facility

So, a nice un-VMSsy non-informal message: why does it not say / how do I find out WHICH facility?

I did on one node
- stop & disable LPD
- stop TCPIP entirely
- start TCPIP
- enable & start LPD

No difference.

On all nodes
- stop & disable LPD
- enable & start LPD

No difference.

I am getting desparate (as are the users and my managers!)

HELP!!!!

Proost.

Have one on me.

jpe

[Volker Halle]

Jan,

exactly on which command do you get those error message ? Di you run the appropriate DCL procedures with SET VERIFY to find out the exact command ? Did you consider to use SET WATCH FILE/CLASS=MAJ after you've isolated the offending command ?

Without being able to exactly see or reproduce the failure, it will be hard for anyone to try to help you.

Volker.

[Volker Halle]

Jan,

after looking around further in TCPIP$*.COM and TCPIP$*.EXE files, I'm getting the impression, that there definitely is no LPD_CLIENT service anymore in the more recent versions of TCPIP (beyond V5.1). The reference in TCPIP$CONFIG.COM just seems to be a leftover.

Could it be, that there is some 'old' procedure or image lying around, which explicitly references LPD_CLIENT ?

Volker.

[Walter Miller_1]

Search through sys$manager:tcpip*.com for LPD_CLIENT. I found TCPIP$V51_CONVERSION.COM stating that LPD and LPD_CLIENT were merged together. It adds LPD to the configuration database, removes LPD_CLIENT from the configuration database and adds LPD to the service database (as needed). See if you have a conversion file on one of the nodes.
I have attached a copy of the V51 com file.

[Jan van den Ende]

Volker,

SET WATCH tell you exactly NOTHING. (rather surprising).

We have robooted one node, and the effect is that I do NOT have the error in the bootstrap (well, LPD was disabled).
Starting LPD gives the error again, and NO LPD queus can be started, because of the same error.

I could not find any old programs or procedures, but I can not be sure of that (how to recognise them?) But then again: it WAS running, and old stuff would have prevented it then as now.

:-( :-(

Proost.

Have one on me.

jpe

[Volker Halle]

Jan,

let me repeat the question: on which command or image do you get the error message ?

Something seems to be looking for a LPD_CLIENT entry in the service database. There is no such thing since TCPIP V5.1 ! Under previous versions, you had LPD/LPR under both the client and server menue in UCX$CONFIG.COM

Volker.

[EdgarZamora]

Jan, can you give us a better idea of the environment (versions, cluster?, common system disk?) I'm thinking that some conversion steps may have been skipped in the past. Maybe the conversion step was only run on one node and you have to run it on the others, etc? Please give us details on the environment so we can analyze this better.

[Volker Halle]

Jan,


I could not find any old programs or procedures, but I can not be sure of that (how to recognise them?)


You're running TCPIP V5.4 ECO 4. Any TCPIP$* file older than around DEC-2003 (that's when TCPIP V5.4 first showed up on the LP CDs) may be 'too old'.

Volker.

[Jan van den Ende]

Volker,

I have been able to get output from SET WATCH.
The error is just after accessing TCPIP$LPD_UTILITIES, which attemps LIBRTL_D56_TV (status 1), then status 910 (??? isn't that file not found?), then TIE$MESSAGES.EXE , and the error display.

Does that tell you anything????

Proost.

Have one on me.

jpe

[Volker Halle]

Jan,

we're getting there...

Is the error on the

$ lpdutilities -f "START_SERVER_QUEUES"
or
$ lpdutilities -f "START_CLIENT_QUEUES"

command ?

Try SET WATCH FILE/CLASS=ALL to get more information.

Does the TCPIP$LPD_UTILITIES.EXE file look o.k. ?

Volker.

[Jan van den Ende]

Volker,

most of those are from 18-sep-2003; only SYS$EXAMPLES holds a lot of older stuff.

Proost.

Have one on me.

jpe

[Jan van den Ende]

Volker,

$ lpdutilities -f "START_SERVER_QUEUES"

seeks and does NOT find TCPIP$LPD.CONF

then accesses TIE$MESSAGES.EXE

and gives the error.

AFAIK that .CONF is optional. It seems we never had it.

Does this help?

Proost.

Have one on me.

jpe

[EdgarZamora]

Jan, could you supply the output of show logical tcpip$lpd* Thanks

[Jan van den Ende]

Edgar,

Two-site cluster, 2 nodes each site.
VMS 7.3-2, TCPIP V5.4 ECO 4
Common, shadowed 3 member system disk.
Most system control files (UAF, QUE, etc) off system disk on one cluster-common shadow set.
TCPIP 5.1 conversion executed on all 5 nodes (one has long since left, conversion was during that replacement period)

Proost.

Have one on me.

jpe

[EdgarZamora]

Jan,

Does this apply to you (especially the last part): (You might want to check your TCCPIP$LPD_SYSTARTUP file)

2.6.3.1 Preserving LPD Behavior (Alpha Systems)
When you install TCP/IP Services over an earlier version of the product, follow
the instructions displayed on your screen to preserve your edits in the LPD
startup and shutdown command procedures.
The following shows a sample screen display.
The following product will be installed to destination:
DEC AXPVMS TCPIP V5.1-9 DISK$ALPHASYS:[VMS$COMMON.]
UCX product already installed.
***********************************************************************
Another version of TCP/IP is installed. You must execute the following
three commands before continuing with this installation:
$ BACKUP SYS$COMMON:[SYSMGR]UCX$LPD_STARTUP.COM; -
SYS$COMMON:[SYSMGR]TCPIP$LPD_STARTUP.COM;
$ BACKUP SYS$COMMON:[SYSMGR]UCX$LPD_SHUTDOWN.COM; -
SYS$COMMON:[SYSMGR]TCPIP$LPD_SHUTDOWN.COM;
$ PRODUCT REMOVE UCX
***********************************************************************
2â 4 Installation, Configuration, and Startup Notes
Installation, Configuration, and Startup Notes
2.6 Removing Prior Versions of this Product
After you follow these instructions and complete the installation of TCP/IP
Services, your site-specific edits to the LPD startup and shutdown files are found
in:
SYS$COMMON:[SYSMGR]TCPIP$LPD_STARTUP.COM_OLD
SYS$COMMON:[SYSMGR]TCPIP$LPD_SHUTDOWN.COM_OLD
Now merge your site-specific edits into:
SYS$COMMON:[SYSMGR]TCPIP$LPD_SYSTARTUP.COM
SYS$COMMON:[SYSMGR]TCPIP$LPD_SYSHUTDOWN.COM
2.6.3.3 Merging Edits (Alpha and VAX Systems)
When you merge edits, do not include the commands to start and stop the queue
UCX$LPD_QUEUE. This queue has been replaced with TCPIP$LPD_QUEUE.
The commands for starting and stopping TCPIP$LPD_QUEUE are in the LPD
startup and shutdown command procedure files.
After you merge the edits, modify the value of the /PROCESSOR qualifier in the
LPD client queue startup commands that you have just appended, replacing
TCPIP$LPD_SMB with UCX$LPD_SMB. For example, enter the following
command:
LSE Command> SUBSTITUTE/ALL "ucx$lpd_smb" "tcpip$lpd_smb"

[Volker Halle]

Jan,

do you find the string 'LPD_CLIENT' or at least '_CLIENT' or 'CLIENT' in either:

SYS$SYSTEM:tcpip$lpd_utilities.exe
SYS$SHARE:TCPIP$LPD_SHR.EXE

in a context, which might be used as a parameter for the STARTFAIL or NOSERVREC error message ?

What does TCPIP SHOW VERS/ALL report for those 2 files ?

When did that system last get booted successfully (i.e. with correctly starting LPD queues) before 2-DEC-2007 ? What happenend since then ?

Volker.

[Wim Van den Wyngaert]

TCP 5.3 eco 2.
TCPIP$CONFIG.COM still has references to LPD_CLIENT.
But the service is not in our ena servi list. I tried to add it and it was accepted. I tried to enable the service and it failed (is not in the services list).

BTW : what a syntax. ucx set conf ena NOservi lpd_client

fwiw

Wim

[Volker Halle]

Jan,

in this thread, you have given 2 error messages:

%TCPIP-E-STARTFAIL, failed to start LPD_CLIENT
%TCPIP-E-NOSERVREC cannot find LPD_CLIENT service database record.

and later

Message number 07648312
= %TCPIP-E-FACNOTSTART, attempt made to run a TCPIP image before starting the facility

I'm assuming that the 2nd message is issued by TCPIP$LPD_UTILITIES, right ? If so, which image/procedure emits the first error message, the one explicitly referencing LPD_CLIENT ?

Volker.

[Jan van den Ende]

Volker,

The STARTFAIL and the NOSERVREC messages are right after one another in executing TCPITP$LPD_STARTUP

>>>
Message number 07648312
= %TCPIP-E-FACNOTSTART, attempt made to run a TCPIP image before starting the facility
<<<

is issued by START/QUEUEU

Proost.

Have one on me.

jpe