[Dspam-user] Dspam not Retraining

[Eric H. <mrr...@ho...>]

Hello all, I have dspam setup with the hash driver, a single merged group including all users, and train pristine set to on. The strange thing that is happening is that only some of my messages are properly retraining. For retraining I am using the command
sed '1,2d' /home/dspam_train/tmp.ooF6h5KhyL | /usr/bin/dspam --user xxx --deliver --class=spam --source=error --debug
I cut  off the first two lines because that is where my script stores the class and user.
Sometimes the output looks like:10150: [02/21/2012 17:46:01] pass-thru args: --deliver10150: [02/21/2012 17:46:01] processing user 22010150: [02/21/2012 17:46:01] uid = 1001, euid = 1001, gid = 1001, egid = 10710150: [02/21/2012 17:46:01] loading preferences for user 22010150: [02/21/2012 17:46:01] default preferences empty. reverting to dspam.conf preferences.10150: [02/21/2012 17:46:01] Loading preferences from dspam.conf10150: [02/21/2012 17:46:01] using /var/spool/dspam/opt-in/local/220.dspam as path10150: [02/21/2012 17:46:01] using /var/spool/dspam/opt-out/local/220.nodspam as path10150: [02/21/2012 17:46:01] sedation level set to: 510150: [02/21/2012 17:46:01] Loading 125 BNR patterns10150: [02/21/2012 17:46:01] Whitelist threshold: 2010150: [02/21/2012 17:46:01] [graham] [1.000000] how+to (2frq, 0s, 8i)10150: [02/21/2012 17:46:01] [burton] [1.000000] how+to (2frq, 0s, 8i)10150: [02/21/2012 17:46:01] [burton] [1.000000] how+to (2frq, 0s, 8i)10150: [02/21/2012 17:46:01] [graham] [1.000000] to+#+#+another (2frq, 0s, 4i)10150: [02/21/2012 17:46:01] [burton] [1.000000] to+#+#+another (2frq, 0s, 4i)10150: [02/21/2012 17:46:01] [burton] [1.000000] to+#+#+another (2frq, 0s, 4i)10150: [02/21/2012 17:46:01] [graham] [1.000000] Subject*electric+bill (1frq, 0s, 3i)10150: [02/21/2012 17:46:01] [burton] [1.000000] Subject*electric+bill (1frq, 0s, 3i)10150: [02/21/2012 17:46:01] [graham] [1.000000] electric+bill (2frq, 0s, 3i)10150: [02/21/2012 17:46:01] [burton] [1.000000] electric+bill (2frq, 0s, 3i)10150: [02/21/2012 17:46:01] [burton] [1.000000] electric+bill (2frq, 0s, 3i)10150: [02/21/2012 17:46:01] [graham] [1.000000] Subject*how+to (1frq, 0s, 3i)10150: [02/21/2012 17:46:01] [burton] [1.000000] Subject*how+to (1frq, 0s, 3i)10150: [02/21/2012 17:46:01] [graham] [1.000000] Â +#+#+The (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] Â +#+#+The (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] percent+#+the (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] percent+#+the (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] all+the (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] all+the (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] the+#+you (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] the+#+you (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] for+#+#+to (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] for+#+#+to (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] the+#+#+#+only (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] the+#+#+#+only (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] he+was (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] he+was (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] Most+of (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] Most+of (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] he+#+#+#+a (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] he+#+#+#+a (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [graham] [1.000000] create+a (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] create+a (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] The+#+#+#+it (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] the+light (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] on+#+#+#+to (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] out+and (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] million+#+#+in (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] who+is (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] under+#+#+#+in (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] the+#+#+#+following (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] [burton] [1.000000] instead+of (1frq, 1s, 2i)10150: [02/21/2012 17:46:01] Graham-Bayesian Probability: 1.000000 Samples: 1510150: [02/21/2012 17:46:01] Burton-Bayesian Probability: 1.000000 Samples: 2710150: [02/21/2012 17:46:01] Result Confidence: 1.0010150: [02/21/2012 17:46:01] libdspam returned probability of 1.00000010150: [02/21/2012 17:46:01] message result: SPAM10150: [02/21/2012 17:46:01] DSPAM Instance Shutdown.  Exit Code: 0
Which means the training worked. But more often the output is:10157: [02/21/2012 17:46:01] pass-thru args: --deliver10157: [02/21/2012 17:46:01] processing user 22010157: [02/21/2012 17:46:01] uid = 1001, euid = 1001, gid = 1001, egid = 10710157: [02/21/2012 17:46:01] loading preferences for user 22010157: [02/21/2012 17:46:01] default preferences empty. reverting to dspam.conf preferences.10157: [02/21/2012 17:46:01] Loading preferences from dspam.conf10157: [02/21/2012 17:46:01] using /var/spool/dspam/opt-in/local/220.dspam as path10157: [02/21/2012 17:46:01] using /var/spool/dspam/opt-out/local/220.nodspam as path10157: [02/21/2012 17:46:01] sedation level set to: 510157: [02/21/2012 17:46:01] DSPAM Instance Shutdown.  Exit Code: 0
Meaning nothing happened. and dspam_stats does not change.
Any ideas on this one? I have read through everything and I can't find a difference between emails that do and do not retrain.
Thanks!-Eric