Milter-greylist

> Any issues running milter-greylist under a pretty heavy load?  The
> servers are dual processor (2.2GHz Xeon) Proliant DL580s with 2GB memory
> each, running RHEL 2.1 ES.  The version of the Sendmail RPM is
> 8.11.6-28.72.

We're running milter-greylist 1.6 on a cluster of FC1/3 servers (dual
processor Dells, 1GB memory) for a user community of about 40,000 and a
greylist.db which has about 600,000 entries.  The CPU isn't breaking a
sweat and memory usage appears fine, even though we run SpamAssassin and
ClamAV on the same servers, but we're running into occasional problems[*]
and are debating whether to enable lazyaw as a possible fix.

In addition to greylisting, we're using RBLs that turn away about
30,000 connections a day.

[*] Before someone with a hammer 'fixed' it by staggering daily
    restarts of the milter, milter-greylist would occasionally die with
    "pthreads can't create process: out of mem".  we also saw "cannot
    sync with peer 1.2.3.4:  out of space (1024 entries queued)" from
    time to time.

--
Dawn Keenan
University of Waterloo, Waterloo ON Canada