Configuring and using DSPAM

[mintra]

Thankyou for that quick reply!

It was my understanding that Zimbra installs it's own mysql server and there shouldn't be any other mysql servers installed.

This is output from whereis mysql

$ whereis mysql
mysql: /usr/bin/mysql /usr/lib/mysql /usr/share/mysql /opt/zimbra/bin/mysql.server /opt/zimbra/bin/mysql /usr/share/man/man1/mysql.1.gz

thanks,
Ian

[marcmac]

You'll find mysql headers and libs in /opt/zimbra/mysql, if you've got zcs installed.

[mintra]

Thanks MarcMac

the configure command finds the folders now, but I still get

configure: error: Required version of libmysqlclient not found

[unilogic]

Its pretty easy to train dspam with a the corpus files from spamassassin.
Download http://dspam.nuclearelephant.com/sou...trainer.tar.gz

Edit publiccorpus.pl as follows on line 50:

$cmd = "/opt/dspam/bin/dspam --user $user --class=$class --source=corpus --mode=teft --feature=chained,noise --stdout < $corpus

Goto http://spamassassin.apache.org/publiccorpus/ and download all the spam and ham files except 20030228_easy_ham_2.tar.bz2 to keep the number of ham and spam files even. Extract these files in the same directory as publiccorpus.pl.

Run: perl publiccorpus.pl zimbra
After it finishes, it will take a good half an hour depending on your cpu power as you're entering thousands of emails into its database,

Run: dspam_clean -p0 zimbra
It will clean up any uneeded or neutral entries in dspam's database. This will also take a good deal of time.

-Ben

[unilogic]

I've also modified zmtrainsa to train dspam. It grabs email in the same fashion as it does for SA training except it sends the messages to dspam for training. Add to /opt/zimbra/bin/zmtrainsa as follows:

if [ $MODE == "spam" ]; then
        DMODE="spam"
elif [ $MODE == "ham" ]; then
        DMODE="innocent"
fi

echo "Training DSPAM"

/usr/bin/fetchmail -n -a -u $USER $FOLDER -m "/opt/dspam/bin/dspam --user zimbra --class=${DMODE} --source=corpus --mode=teft --feature=chained,noise --stdout" ${SERVER}

echo "You should run: dspam_clean -p0 zimbra to clean up useless dspam entries"

Make sure this follows the added code

rm -f ${zimbra_home}/.fetchmailrc

It should work, It'll train SA and DSPAM. Let me know if anyone has problems.

-Ben

[marcmac]

Might be faster to use on fetchmail command to dump to a temp file, then send that file through both SA and DSPAM - cuts down on the network traffic.

[vcolombo]

Ok, I followed the instructions and everything seems to have gone well. At first I was getting errors when running dspam_stats but I seem to have gotten that resolved. My question now is how can I verify that dspam is running and functioning correctly?

Thanks

[phoenix]

Have a look in the Zimbra logs and see if DSPAM is recognised when you start Zimbra, you should also have a look at the 'original' email and see if DSPAM has put any headers in there. You'll also have some logs in the DSPAM directory, have a look at them and see if they show anything.

You'll also get some stats if you run dspam_stats -H as root.

[vcolombo]

Ok, so after doing a cat /opt/zimbra/log/* | grep dspam I don't see any results, although running dspam_stats -H does yield the following:

zimbra:
TS True Positives: 0
TI True Negatives: 0
IM False Positives: 0
SM False Negatives: 0
SC Spam Corpusfed: 2398
IC Innocent Corpusfed: 5551
TL Training Left: 0
SR Spam Catch Rate: 100.00%
IR Innocent Catch Rate: 100.00%
OR Overall Rate/Accuracy: 100.00%

So at this point I'm assuming it's working.

Has anyone managed to get the dspam web interface installed on their Zimbra server? It would be nice to have access to the graphs and all.

Thanks again for the great tutorial and feedback. I'm truly impressed by both the Zimbra product and the Zimbra community.

-Vince

[marcmac]

You probably want /var/log/zimbra.log, not /opt/zimbra/log/*