BIND Resolver keeps disabling in UCX

[IST Windsor]

I have configured UCX to use our corporate DNS and enabled the BIND resolver. In UCX the resolver initially shows as enabled; All is well and host names can be resolved:

UCX> show name

BIND Resolver Parameters

Local domain: abglobal.com

System

State: Started, Enabled

Transport: UDP
Domain: corp.abglobal.com
Retry: 4
Timeout: 4
Servers: WNDDCP01, WNDDCP02

UCX> show host ABBAPSFND01

BIND database

Server: 159.222.182.100 WNDDCP01

Host address Host name

103.22.28.11 ABBAPSFND01.corp.abglobal.com

A few minutes later the state will go to disabled and host names will stop being resolved:

UCX> show name

BIND Resolver Parameters

Local domain: abglobal.com

System

State: Started, Disabled

Transport: UDP
Domain: corp.abglobal.com
Retry: 4
Timeout: 4
Servers: WNDDCP01, WNDDCP02

UCX> show host ABBAPSFND01
%UCX-E-HOSTERROR, Error processing HOST request
-UCX-W-NORECORD, Information not found
-RMS-E-RNF, record not found

What is even more odd is that some minutes later it will enable itself again.

I can find no reason for the state to change and no logfile entries which give any clues. The same thing happens on two unconnected systems, one Alpha and one VAX, running VMS 7.1-2.

Any pointers appreciated, thanks.

[Volker Halle]

Hi,

this sounds pretty unusual...

The state of the BIND resolver is kept in a couple of logical names:

$ SHOW LOG UCX$BIND*

You could analyze the state with a little DCL procedure to at least get an idea how often and when it is changing:

$ x=F$TRNLNM("UCX$BIND_STATE")
$ x1=F$CVSI(0,32,x)
$ x2=F$CVSI(32,32,x)

The state is maintained in 2 longwords, symbols x1 and x2 represent those longwords and you could check in a little loop, if and when there values are changing.

x1 will changed from 3 to 2, if system-wide BIND resolver state is disabled.

Only SYS$SHARE:UCX$ACCESS_SHR seems to access those logicals.

You could also try to experiment with NSLOOKUP:

$ MC UCX$NSLOOKUP
>set debug
>set d2
>ABBAPSFND01
..
>exit

And also try tracing, if you spot anything unusual - maybe at about the same time, when the BIND resolver state changes to disabled:

$ tcpiptrace WNDDCP01 /port=(remote=53)

Do you have other OpenVMS systems connected to those DNS server, where this does NOT happen ? Are you running the most recent UCX patch level on those systems ?

Hope this helps to shed some light on this problem.

Volker.

[IST Windsor]

Thanks Volker, I used NSLOOKUP to determine that I can lookup the name in one window even when in another, at the same time, I cannot ping the node:

> nabva1
Server: wnddcp01.corp.abglobal.com
Address: 159.222.180.100

Non-authoritative answer:
Name: nabva1.abglobal.com
Address: 159.222.88.89

$ ucx ping nabva1
%UCX-E-GETHST, Error in getting host name
%RMS-E-RNF, record not found
$ ucx show host nabva1 /nolocal
%UCX-E-HOSTERROR, Error processing HOST request
-UCX-E-BINDDISABLE, BIND Resolver is disabled
-SYSTEM-S-NORMAL, normal successful completion

TCPIPTRACE showed nothing apart from the pings going outand the replies coming back, there were no other messages when BIND disabled.

I think this must be a image/version/patch problem, it's too wierd to be anything else!

[IST Windsor]

I have been pointed at a bit of code buried in a housekeeping batch procedure on our systems which manually stops and starts the BIND Resolver in order to work around a long-forgotten issue with SMTP mail. This is definitely the cause of the problem.

Sorry to have bothered everyone with it; thanks to Volker for helping out, at least I learned how to NSLOOKUP and TCPIPTRACE.