TCPIP Pause At Startup

[Wim Van den Wyngaert]

Got it.

ucx start mail will go in hang when DNS is enabled. If you disable it, everything goes fine.

Wim

[Fekko Stubbe]

You will also experience (a long) delay if you have proxies defined. TCPIP will try to translate all proxy hostname entries, and this may take a lot of time (if you have a lot of proxies)

[Robert Atkinson]

> It's the startup of SMTP that is delaying everything.

Thought that's what I'd said??

This is the text I wrote for my DR plan :-

"If the DNS server is not available when the system is booted, the boot will hang trying to start SMTP for 10 minutes until it reaches the timeout. To prevent this, action the following commands.

@SYS$MANAGER:TCPIP$CONFIG
Select 2 â Client Components
Select 6 â SMTP
Select 1 â Disable Service"

[Wim Van den Wyngaert]

Sorry Robert. You were right.

Still curious if there is a way to decrease the time to less than 1 minute.

Wim

[Robert Atkinson]

Only by using the method outlined above, or hacking the TCPIP$STARTUP in some way to detect the presence of the DNS server (which is probably not possible, as you need IP up to do that).

A lot of other people have used the SYSGEN user startup parameters to load information into, then interrogate at boot time. Perhaps you could try that?

[Wim Van den Wyngaert]

You can disable smtp and enable it later in the startup but via a submit.

But I'm not going to prevent the hang : I'm not in a hurry to get my node up without network.

Wim

[Dean McGorrill]

Wim,
tcpip, like decnet, were pretty good about having dials either through commands or logicals to change stuff like timers. esp one of 10 minutes. I'd be surprized if
there isn't one. if theres someone inside
hp in here, they could ask in the tcpip notes conference (or its equiv) or look
in the tcpip sources.

Dean

[Robert Atkinson]

Not to say one doesn't exist, but I had a good hunt around at the time and couldn't find it.

By the way, if you want to see if you can reduce the timeout, simply issuing a '$ NSLOOKUP' command will have exactly the same effect as the boot hang, taking a number of minutes to realise the DNS server isn't responding.

We have 3 DNS servers configured, which is probably why our timeout takes 10 minutes.

Rob.

[labadie_1]

I have done
$ @sys$startup:tcpip$define_commands
$ sysconfig -q inet
$ sysconfig -q socket
$ sysconfig -q net

and I have not found a parameter with a value of 600 or 6000 (if the unit is 1/10 second)

of course this proves nothing, as the unit for this timeout may be any value.

May be it is hard coded and it would need to patch some cell before and after the start of Tcpip (or copy before and after 2 versions of the .exe holding the timeout value).

[Wim Van den Wyngaert]

set config nam/retry=1/timeout=1
(were both 4, for 2 name servers in /serv)

Startup time of ucx went from 4min27sec to 19 sec. Normal startup takes 12 sec (very old slow as500). So 1 retry takes 8 sec for 2 nodes.

Retry=2 took 27 sec.
retry=3 took 43 sec.
retry=0 took 1min15sec ???

So 4 seconds per retry per node.

The arp timeout is about 4 sec. So it's arp that's to blame. Note that I traced IP in the beginning : this was a fault. Always start with tracing ARP.

And I think ARP timeout is hardcoded in tcpip.

Wim