Hi Oliver,
Try these subtle enhancements to your transport - I use the same tools
but on RedHat:
[ snip ]
> home_directory = "/tmp"
> current_directory = "/tmp"
> user = Debian-exim
> group = Debian-exim
> return_fail_output
> message_prefix =
> message_suffix =
> log_output = true
I've always had the following in addition to the above:
batch_max = 5
return_path_add = false
This helps SA from getting bogged down - the batch_max -- under way too
heavy load, SA seems (to me!) to not always perform correctly. Adding a
lot of RAM could help - spamd is a memory pig. Taking a glance, I have
three spamd procs each using a SIZE 116meg and RSS 102meg. Oi veh.
Such a large batch condition will happen when you have an alias expanded
to a bucketload of final delivery addresses - ie a project/staff alias.
Adjust the value to your machinery power of course. You could also try
rearranging your routers so that SA is called *before* final alias
expansion, so that it's scanned only once.
-te
--
Troy Engel | Systems Engineer
Fluid, Inc |
http://www.fluid.com
