New to PMG and have some questions

[sagelike]

Hello,

I am new to PMG and have some questions:

1. If I Delete or Blacklist an email, is this action shared with Proxmox in anyway so that it learns?
2. If I whitelist and email, is this action shared with Proxmox in anyway so that it learns? ?
3. What does Backup actually backup?
4. If, for example, the server broke and I had to create a new server from scratch, will the backup restore PMG fully to its previous state with all configuration settings, Relay domains, Transports, etc?
5. Is there a way to create and manage backups from the GUI?

Thanks, appreciate your help with this.
G

 

[Stoiko Ivanov]

most questions are answered in our reference documentation:
https://pmg.proxmox.com/pmg-docs/

1. If I Delete or Blacklist an email, is this action shared with Proxmox in anyway so that it learns?
2. If I whitelist and email, is this action shared with Proxmox in anyway so that it learns? ?

which black/whitelist do you mean - PMG has 3 levels of black/whitelist:
* mailproxy whitelist (to skip greylisting/spf and other checks during SMTP dialog) - this is configured in the GUI and PMG of course saves the information
* Blacklist/Whitelist in the rule-system - also configured in the GUI
* user black/whitelist - which is where users can change a way a particular mail is treated - this is meant that a user can decide that a mail is directly delivered to them instead of put in quarantine - this is saved for this users' mailbox only

3. What does Backup actually backup?

see https://pmg.proxmox.com/pmg-docs/pmgbackup.1.html

4. If, for example, the server broke and I had to create a new server from scratch, will the backup restore PMG fully to its previous state with all configuration settings, Relay domains, Transports, etc?

all pmg-specific configs are restored - you need to make sure to configure/backup the network settings and the hostname

5. Is there a way to create and manage backups from the GUI?

yes - GUI->Configuration->Backup/Restore

I hope this helps!

 

[sagelike]

Thanks Stoiko. That helps!

For questions #1 and 2, I was thinking about user Whitelist and Blacklist via the Daily Spam Report.

Does Proxmox collect these actions centrally and add this to its Spam detection rules in some way? I'm not thinking of server level Bayesian learning specifically but whether Proxmox adds this info to its rule sets.

Thanks again for your help.

 

[Stoiko Ivanov]

For questions #1 and 2, I was thinking about user Whitelist and Blacklist via the Daily Spam Report.

this then corresponds to the user black/whitelist:

* user black/whitelist - which is where users can change a way a particular mail is treated - this is meant that a user can decide that a mail is directly delivered to them instead of put in quarantine - this is saved for this users' mailbox only

so no it's not centralized and it should not be - because this is the level where individual users decide what to do with a mail for them - it only changes the decision between quarantine, direct delivery or dropping the mail for them. E.g. if one of your users want to have the newsletter of their favorite lunch-place in their inbox, while the others say - they never read it and want it dropped.

As for bayesian learning - PMG does learn from spam/ham if you enable Bayesian filtering - but only for very high scores (PMG is using SpamAssassin and its Bayes module for this)


Hope that explains it!

 

[sagelike]

One last question. Is Bayesian filtering enabled in: Configuration > Spam Detection > Options > Use Bayesian filters (it is already set to Yes)

Thanks again.

 

[Stoiko Ivanov]

One last question. Is Bayesian filtering enabled in: Configuration > Spam Detection > Options > Use Bayesian filters (it is already set to Yes)

yes - but that amounts to setting 'use_bayes 1' in your spamassassin configuration.
this usually implies autolearning as used by Spamassassin (which needs a few emails with very high scores (or low negative scores) to start working).

some users train their bayes-filter with their sample mails (but there again make sure to read the docs and afterwards keep an eye on the logs for potential misclassification due to bayes)

check the spamassassin documentation on bayes filtering:

https://cwiki.apache.org/confluence/display/SPAMASSASSIN/BayesInSpamAssassin
https://cwiki.apache.org/confluence/display/SPAMASSASSIN/BayesFaq
https://cwiki.apache.org/confluence/display/SPAMASSASSIN/BayesNotWorking

I hope this helps!