1 |
Patrick Marquetecken wrote: |
2 |
> After changing the hdparm parameters and in main.cf |
3 |
> default_destination_concurrency_limit to 20 the speed whent up and i |
4 |
> send now 100 email in seconds. |
5 |
> |
6 |
|
7 |
I'd bump the limit up to 100 or so if the box can handle it. If it |
8 |
can't I'd suggest increasing the RAM. As you've likely noticed disk I/O |
9 |
is going to be your main bottle neck. You can do interesting things like |
10 |
adding a second drive and running another instance of Postfix on it. You |
11 |
can also turn off all logging, but for 10k you shouldn't need to do |
12 |
anything creative. |
13 |
|
14 |
If you've never sent an email to your customers you're going to have a |
15 |
lot of bounces. It's just one of those things. From past experiences I'd |
16 |
consider myself lucky to get 80% working email addresses. This brings me |
17 |
to my main point, bounce handling. Bounce handling is going to be |
18 |
expensive in system resources initially, but your return is pretty quick. |
19 |
|
20 |
Proper bounce handling should update you db with a marking of not |
21 |
active, etc for addresses that bounce. Also if it bounces, it's dead. |
22 |
I've seen some people wait for 5 bounces before marking an address dead. |
23 |
That's just dumb. I'd go maybe as high as 2 bounces, this ain't 1996 and |
24 |
mail pretty much just works now. All bounce handling should be done |
25 |
before the next mailing, so you can run with a cleaner lists. You should |
26 |
also run a few reports on you logs to see if anyone is marking you as |
27 |
spam. Again it's one of those things. Some idiot somewhere will complain |
28 |
and you'll either be blocked locally or end up on a list no matter how |
29 |
in the right you are. Best you're prepared to deal with it. |
30 |
|
31 |
kashani |
32 |
|
33 |
|
34 |
-- |
35 |
gentoo-user@g.o mailing list |