[exim] Host-based retrying problem

Hello guys,

Given that Exim uses host-based retrying, I bump into the following issue.

I have a front-end SMTP server (A) running Exim, dedicated mainly to virus scanning, which then routes cleaned messages to the server holding mailboxes (B) running Sendmail.

incoming connections ---> A (virus scanning) -----> B

This is quite a heavy traffic system, an average of 4 incoming connections per second on A.

The problem is that when A gets a host error from B for some reason, all messages destined to B are queued. And when retry time has come, A sends all messages to B at once, whose load average rises, so leading Sendmail to refuse connections. So emails get queued on A ....

When Sendmail runs on A, everything is fine. I guess that's because it uses message-based retrying so that messages are re-sent 'on the flow'.

Any idea ?

Thanks in advance.

Christian