Author: W B Hacker Date: To: exim users Subject: Re: [exim] Callouts that don't immediately work
Jaap Winius wrote: > Hi folks,
>
> After recently configuring a new mail server (exim4-daemon-heavy 4.72-6
> on Debian squeeze) with which I was quit satisfied, I was informed that
> the system mail name (the domain name used to qualify mail addresses
> without a domain name) had to be changed. After adding the new domain
> name I ran some tests and noticed that the callouts from other servers
> receiving from this one were not succeeding. No idea why. The callout
> still worked for the old domain name. Using telnet, I could run the
> callout procedure manually for addresses using the new domain name, but
> that worked just fine.
>
> Frustrated, I backed off and thought about it for a while. Then I ran
> some more tests a few hours later... which all completed without any
> problems. As if nothing had ever been the matter.
>
> Is this normal behavior? Any explanations?
>
> Thanks,
>
> Jaap
>
'Typically' on a whole-globe basis, nameservers need as much as three
days before changes have propagated to 'near as dammit' all of them.
Ordinarily, one sets the TTL's quite short in the period leading up to a
planned change. But this is not always wise for an MTA, so there's a
trade-off.
If someone else was looking after that part, you may not even be sure if
it was started before you made your changes - or well after.
Nor would one always be quite sure when other MTA's updated locally
cached information (if any...).
We generally test as we go. See traceroute.org and looking-glass servers.
Hard to change the equation. Too many diverse players. But one can at
least be aware - make a few interim compromises to reduce the impact.