Long queue processing

Question

Hello,
first I'll describe the situation:
- we have test environment which sends e-mails to non existing e-mail addresses
- we have catch_all mailbox (using catchallagent.codeplex.com )
- on that mailbox we have rule for incoming mail that redirects all mail with sender containing '@domain.com' to another mailbox and deletes the e-mail (put it in trash)
- on both mailboxes we have retention policy which clears trash after 1 day (no possbility to set it to shorter)
- volume of these test e-mail is 3000 / day (one shot is from 10 to 20 e-mails at the same time)
- we have Forefront for Exchange
- both mailboxes are on the same database and it's separate database from rest of company e-mails
- hardware wise it's 2 cores and 16GB ram, only exchange and forefront on this virtual machine
Now the problem is that queue often hangs on redirecting the e-mails (from catch_all mailbox to second one) and takes very long time to process, sometimes errors also if you try to access said mailboxes through OWA it's impossible (timeout). Which in the
end leads sometimes to hanging of the queue and requires Transport Service restart.
The error on messages is 4.3.2 432 STOREDRV.Deliver; recipient thread limit exceeded.
We applied the fix suggested, adding:
<add key="RecipientThreadLimit" value="2" />
<add key="MaxMailboxDeliveryPerMdbConnections" value="3" />
To EdgeTransport.exe.config but it didn't help. We also tried increasing the values by +1 to 3 and 4.
Restarting Transport Service is temporary solution.
Any advice is very welcome on this problem.
Best regards,
Wojciech Iwański

Answer 1

It looks like it hangs on one specific one as Active for some time and then tries to proceed other ones but all get fail with error 4.3.2
432 STOREDRV.Deliver; recipient thread limit exceeded.
We've noticed that the problem mainly starts when the mailbox has lot of items (like 2000-3000 daily mail in trash before deletion), we've written simple script that connects to the mailbox and cleans e-mails after 3h and so far we've not noticed a queue
hangs, but it's work around for me not really good solution.