How do you flush tcpip?

[Mike Smith_33]

I am running VMS 7.3-2 with tcpip 5.4 ECO 5. As of last month, all required patches installed. I have found that tcpip has a nasty habit of remembering when it encounters errors, I do not know how long it remembers.

On Windows type systems there is a flush dns type command that tells it to actually try the connection again. Does anyone know how to do the same thing from tcpip on VMS?

[Wim Van den Wyngaert]

Don't use bind but try

$ tcpip help set name_service /init

Wim

[Mike Smith_33]

Wim here is what I find in the help

Reloads all BIND server databases and the BIND configuration
file. Either TCPIP$ETC:RNDC.CONF or TCPIP$ETC:RNDC.KEY must be
set up to allow for secure communication between the user and
the BIND server. For more information, refer to the HP TCP/IP
Services for OpenVMS Management manual.


From reading this, it looks similar to a flush. Can anyone verify who has actually done this?

[Arch_Muthiah]

Mike,

I have never had this problem, but try this

$ RUN SYS$SYSTEM:NCL

--- To flush entire naming cache....

NCL> flush session control naming cache entry "*"

---remove one or more entries from the in-mem local naming cache use this cmnd

NCL> flush [node node id] session control naming cache entry "node name"

--- to flush single entries...
NCL> flush session control naming cache entry "IP$nnn.nnn.nnn.nnn"


Archunan

[Arch_Muthiah]

Mike,

For more info, please see this FAQ under
15.3.2 What does "failure on back translate address request"
mean?

http://www.cs.uu.nl/wais/html/na-dir/dec-faq/vms/part10.html


Archunan

[Jan van den Ende]

Archunan,

NCL relates to DECnet,
this question is about TCPIP.

Apart from both being networking, they have CERY little in common!
(DECnet CAN be configured to use IP ( _NOT_ TCP/IP ! ) as carrier, but that is all).

just my EUR 0.02

Proost.

Have one on me.

jpe

[Ian Miller.]

Mike,
can you say in detail what flush dns does on Windows?

[Steven Schweda]

> [...] tcpip has a nasty habit of
> remembering when it encounters errors, [...]

What sort of errors?

A "flush dns type command" might clear a DNS
cache, but only a DNS user can "actually try
the connection again".

> $ RUN SYS$SYSTEM:NCL
> [...]

Thanks, AM. That was about as useful as
usual.

You can always try:

@ sys$manager:tcpip$bind_shutdown.com
@ sys$manager:tcpip$bind_startup.com

The TCPIP documentation may say more about
flushing the DNS cache, but you can probably
read that about as well as anyone else.

[Mike Smith_33]

Wim, I will assign your points once I know whether that works or not.

Thanks all for the replies.

To answer the question about a flush dns command. If you look up the help in msdos on ipconfig it has an option called flushdns which PUrges the DNS Resolver cache

Sometimes when attempting to access an ip address that has a problem, even after fixing the problem you can't immediately get to it because somewhere in there it gets cached and the system just reports back the previous error. By doing a flushdns you force it to go out and actually check because you clean out the cache.

I recently had an issue with a node that was coming across tcpip with the wrong name attempting to access nfs exported drives. The authentication was based on what the nodename should have been thus nfs would not allow it to talk to the drives. I tried a full restart of tcpip with no luck.

Today we had a situation with all of our VMS telnet queues. The problem was a network box that had failed but even after the network guy fixed that VMS would not talk to the queues because it was still cached that there was a problem. We had to stop and restart hundreds of queues to get this working again.

I hope this helps and once again thanks for the replies.

[Wim Van den Wyngaert]

Also consider clearing the ARP cache. I never tested it but it might not be cleaned when a name/IP address is changed while it is in use.

Wim@home

[Arch_Muthiah]

Steven,

> Archunan,
> NCL relates to DECnet,
> this question is about TCPIP.

See how professionally Mr.Jan let me know my mistake?

and see how you are?
> $ RUN SYS$SYSTEM:NCL
> [...]
> Thanks, AM. That was about as useful as
> usual.

What happened to you Steve?. May I know Why do you keep made at me?

If you look back, sometime even after you blaming my suggestions, my suggestions become the correct one and your answers has been proved useless with zero points. Even in those situation I avoided rising your useless/irresponsible blaming.

I knew we both are nothing in this ITRC, understand friend we are just kids in this forum. There are so many Jaambavans (Innntttellectuals) in this forum like Ian, Procura, Uwe Zessin, Volker, Jan, Wim, Clay Stephenson, Steven E. Protter, James R. Ferguson, etc, etc.

Come on friend Steve. "When we took birth we did not bring anything, similarly when we end our life we are not going to take anything". life is very short, nothing permanent, you and me are just invisible in this Universe. take it easy friend if any of my posting hurted you.

Cheers
Archunan

[Michael Yu_3]

Hi Mike,

The problem might not be related to the bind resolver. It might be caused by the ICMP redirects when the "network box" failed. Check the routing table and see if there are dynamically learnt routes.

Use TCPtrace or TCPdump to see if any connection request is sent from the OpenVMS box.

Thanks and regards.

Michael

[Mike Smith_33]

I thank you all for your responses. I have found out that the problem was not within VMS at all. The dns guys made several changes to the secondary dns server to which I point, two of which impacted me. One was to turn off zone forwarding and the other was to turn off reverse name lookup. The secondary dns server was also configured to go directly to root hints when it could not resolve a host name instead of to another internal server. It resolved one of my unix boxes to an outside address and the other to an ip address. A few minutes after these changes were reset, nfs began working like a champ again.

[Stefan Risch]

...try $rndc flush :-)