Zied Fakhfakh <[email protected]> wrote:

> Only the Error message received at the client side (Zimbra webmail actually) 
> does not reflect the data I filled in, is it expected ?

How on earth should I know - I don't know what you "filled in" or where, or 
what your system configuration is, or what error message you got, or ...

Ideally you'd specify :
What you did configuration wise
What you did to test it
What you expected to happen/get
What actually happened/you got

But throwing a few guesses in ...
If you configured a policy to rate limit a client, and then you tried sending 
messages above that rate, then I'd expect the client to give you an error 
message. That error message is client specific, and may or may not include any 
detail of what the mail server said to it - unfortunately quite a few give no 
more detail than "it didn't work" !

The message may well be sent to some of the recipients - that is defined by 
your mail server config. For example, before I got all this sorted out, I had 
Postfix set to allow up to 50 recipients per message, but only 3 errors before 
it aborted reception. Because of the way Policyd does rate limiting, it would 
count the accepted recipients as delivered and update the counters accordingly 
- but postfix would terminate it after 3 fails from Policyd. The result was 
that a customer we had that used us for a relay for mailshots would get 
permanently stuck - each delivery attempt would use up the few recipients 
allowed before hitting the limit, but then get rejected and stuck back in the 
queue. There was never any time for enough quota to develop to allow sending of 
a single 50 recipient message and so nothing went through.
I had to relax the error limit (up to 50). That way, the initial recipients 
would get accepted - and if the quota was hit then the rest of them would get 
rejected, leading to the message getting out to a few recipients at a time at 
the allowed rate. It would go out to all recipients eventually.
I have put in a feature request for recipients to be updated after message 
delivery (like message size is done). That would mean the customer going 
over-quota, but then being blocked completely for a short time until the 
counters bled down.

In your case, using a mail client rather than a server, if (say) the user sent 
a message to 20 recipients when there was only 10 recipients worth of quota 
available - the client would get an error, but 10 of the recipients would get 
the message. If the user then resends the message, it may go to those who 
previously got it, but unless the quota has reset enough, it may well still 
fail again and some recipients will not get it. If the client is deterministic 
in the order of recipients then it will always be the same ones that do get the 
message (those at the start of the list) and don't get the message (those at 
the end of the list).


_______________________________________________
Users mailing list
[email protected]
http://lists.policyd.org/mailman/listinfo/users_lists.policyd.org

Reply via email to