1 |
On Monday 24 January 2011 01:22:09 kashani wrote: |
2 |
> On 1/23/2011 4:26 PM, Alan McKinnon wrote: |
3 |
> > Apparently, though unproven, at 02:02 on Monday 24 January 2011, kashani |
4 |
> > did |
5 |
> > |
6 |
> > opine thusly: |
7 |
> >> On 1/23/2011 12:20 PM, Alan McKinnon wrote: |
8 |
> >>> It manages it's own queues beautifully. But, and this makes me sad, it |
9 |
> >>> doesn't really want *me* to manage it's queues. Border controls are |
10 |
> >>> hard, and finding the 1,000 mails some idiot with a Windows bot just |
11 |
> >>> sent, and deleting them, is really hard. |
12 |
> >>> |
13 |
> >>> I'm redesigning our mail setup at work,a nd I'm going to do it with |
14 |
> >>> exim *and* Postfix. Exim is the front end I can see, work with, and |
15 |
> >>> manage. Exim sends on to Postfix as fast as it can, and Postfix |
16 |
> >>> transparently relays to recipient. I get best of both worlds :-) |
17 |
> >>> |
18 |
> >> I can't say I've ever needed anything more than mailq | grep |awk | |
19 |
> >> |
20 |
> >> postsuper -d - in order to delete mail from the Postfix queues. What |
21 |
> >> sort of things are your trying to do other than delete a lot of spam or |
22 |
> >> bounces? |
23 |
> > |
24 |
> > First, our internal mail system deals with about 3,000,000 mails a day |
25 |
> > Mon-Thu so grep | postsuper is a tad inadequate, even if just on the |
26 |
> > basis of volume |
27 |
> > |
28 |
> > The basic tools are fine as long as you understand what they are dealing |
29 |
> > with - raw text. As soon as you run mailq you have text, you no longer |
30 |
> > have intelligence about what that text means. So you need lots of |
31 |
> > grep-fu. |
32 |
> > |
33 |
> > I can't control what the users mail out, sometimes they have automated |
34 |
> > systems that do silly things like send 10,000 notifications an hour to |
35 |
> > an SMS gateway when they cocked up Nagios. Finding the dodgy ones is no |
36 |
> > fun when there's a lot of perfectly valid ones in the mix too, and grep |
37 |
> > doesn't help much other than blindly selecting text matches. |
38 |
> > |
39 |
> > There's lots more examples, but they all follow a similar theme. |
40 |
> |
41 |
> Thanks for the extra detail, I found what you're describing very |
42 |
> interesting. I've never dealt with Postfix with more than a couple |
43 |
> hundred internal users and more often as spam our customers system. |
44 |
> Other than the occasional Nagios blasts I haven't had to deal with much |
45 |
> of this. |
46 |
> In regards to controlling what users send is it feasible to use a |
47 |
> policy server for rate limiting them? The ability to use an extra lookup |
48 |
> service to decide whether to access main, filter it, allow relay, etc is |
49 |
> one of the things I think Postfix does well. However I suspect the |
50 |
> management and hand holding of a rate limit system would create more |
51 |
> overhead than cleaning out the queue periodically. |
52 |
|
53 |
[Off-topic] Can't you set up nagios to only send out a single alert when a |
54 |
monitored variable goes down - can't remember the parameter off hand but |
55 |
that's what I did when the default nagios setting proved to be too trigger |
56 |
happy for the users' needs. |
57 |
|
58 |
-- |
59 |
Regards, |
60 |
Mick |