Greetings all. We recently started experiencing email delays on our Zimbra installation. Upon further analysis, we found that the amount of incoming spam has tripled in a little over a month. We believe it is this increase in spam that is causing our server to be delayed. We've been trying to free up as much disk I/O and processor/memory cycles as possible on the Zimbra server to help combat the delay, but it has only helped marginally.
In the meantime, I read through the
Postfix Bottleneck Analysis document, and I've found that a ton of the messages on my server are waiting in the
maildrop queue. maildrop is for mails submitted via sendmail. On the document, it specifically states that "High volume sites must avoid using content filters that reinject scanned mail via Postfix sendmail(1) and postdrop(1)." From what I can tell, it looks like Zimbra is doing exactly that.
The maildrop queue is handled by a single thread process: pickup. I reviewed the
pickup man page, but didn't see any advice as to how I can flush that queue or speed up the delivery process from that queue. Does anyone have any tips or reference materials that I can review for more information?
Bugzilla
Wiki
Source
Logjam in the maildrop queue 
Resolution? 
Linear Mode
