[rt-users] Email delay on ticket creation

[Bill Cole]

Derek Cunningham wrote, On 3/12/09 8:30 AM:
> Hi Bill,
> 
> I really appreciate your time and helpful answers.  They have certainly
> pointed me in the right direction, and I'll update the list when I have it
> solved.
> 
> It's been a bit of a challenge to make this run on Mac OS X, all kinds of
> different preinstalled software points to that it 'should' work easily, but
> time and time again I find myself fighting against what should be simple
> config changes that have been harder to solve than I though it should be.

It is useful to keep in mind that Apple acts strategically, even when it 
looks like they are just being sloppy. MacOS X *Server* is a premium-priced 
product and has very few truly secret differences from the standard version 
of MacOS X that comes with a personal Mac, it is mostly a matter of 
configuration; the desktop version of the system is aggressively configured 
towards use as a personal computer that is almost exclusively used through 
the GUI by one person at a time. The Postfix config is an example of this, 
and if you want Postfix on a Mac to behave like a normal server instance of 
Postfix, you should either replace it altogether or at least fix the lack of 
wakeup settings in master.cf. That may also require you to watch out for 
Apple regressing the change in future updates, but it won't break anything 
that already works.

> At least I have some much more focused searches to make about the postfix
> config.  I will try changing RT's $MailCommand and $SendmailArguments first
> and see if that makes the difference.

Well, despite my digression on that option I think that is a second-best 
approach. It is something one might try if one does not have administrative 
control of the mail subsystem or has some need to keep the no-wakeup model. 
The 'sendmailpipe' option for $MailCommand is a much more efficient 
approach, and I wouldn't suggest switching to 'sendmail' on a high-volume RT 
machine. I'm not even sure that change will solve the queue delay problem 
completely. My hypothesis is that it will avoid the problem by serializing 
the queueing of messages for multiple recipients so that the Postfix 
'pickup' process can't win a parallel race with a 'sendmail -oi -t' process 
that has to parse multiple recipients out of a Bcc header.