For a while now, I have been using dspam as a milter strictly to detect spam and mark up the headers of my emails, similar to spamassassin (actually, I have it sitting after spamassassin in the local delivery chain). Emails are not quarantined. The web interface is not used at all. I let the mail client do the work of detecting spam headers and sorting the messages into proper folders (Spam,Trash,etc). False positives/negatives are sent to a designated [EMAIL PROTECTED] or [EMAIL PROTECTED] then piped to the script for retraining. Dspam feeds off mysql for token data, I don't see it being a problem to cluster, however the .stats and .log files are written locally. Not sure if they are vital for dspam to function but the dspam/log and dspam/data directory trees would have to be NFS'd.
~Rolan On 2/22/07, Ricardo Kleemann <[EMAIL PROTECTED]> wrote:
But how can the quarantine delivery be "bypassed"? By default dspam delivers quarantine into its own mbox file, how would that be destined elsewhere? ----- Original Message ----- From: "Samuel Clough" <[EMAIL PROTECTED]> To: <[email protected]> Sent: Thursday, February 22, 2007 4:13 AM Subject: Re: [dspam-users] clustered setup > > That's similar to what I was thinking. If you have the spam delivered to > a user's spam folder in their mailbox rather than put it in quarantine, I > would think you could cluster them, although I haven't done it, so take my > advice with a grain of salt. > > On Thu, 22 Feb 2007 04:14:07 +0100, Tony Earnshaw <[EMAIL PROTECTED]> wrote: >> Todd S. Florman wrote, on 21. feb 2007 23:11: >> >> [...] >> >>> I completely forgot about the mbox/nfs situation, but you are most >>> certainly correct. Does anyone happen to have any innovative ideas >>> about per user quarantines in a clustered sort of setup, or am I stuck >>> using client/server mode with libdspam processing only occurring on a >>> single server. >>> >>> If anyone is running dspam across multiple servers, I would be >>> interested in knowing what you did or concessions you had to make to get >>> a working setup. >> >> We don't run our dspam (or IMAP) setup on shared servers, but we do run >> an alternative IMAP quarantine setup to dspam's native CGI that would be >> suitable for NFS, just as IMAP is suitable for NFS *if* run on a >> dependable backend such as Netgear or NFS 4 (this last has been >> discussed often on the Courier IMAP mailing list). >> >> Users drag 'n drop (Evolution, Thunderbird, Outlook[Express]) or move >> (SquirrelMail) misjudged messages to a "misjudged" folder and a cron job >> reclassifies the messages every hour and moves the reclassified message >> to an appropriate folder (INBOX or INBOX.Spam). The cron job also >> mangles the dspam header so that the message can't be submitted a second >> time. >> >> This is not an answer to your original question, it is an alternative >> way of doing things ;) >> >> --Tonni >> >> -- >> Tony Earnshaw >> Email: tonni at hetnet dot nl >> >> > > > !DSPAM:65534,45dd89bd321548992556831! > >
