Permanently mark email as not junk

Hi everyone, i’m receive email with regularity from my nethserver. Some email are marked as spam. i’am marked this email wiht “not junk” but the system continued marked “spam”.


Hi Francesco

By now you should know: We need more info…

Are you using SoGo, WebTop or just IMAP to a Client?
What client are you using?

As a Tip: If using Roundcube, and set there as Non-Spam, it’s respected by other Clients too. It needs to be moved from Spam to normal Inbox for some clients…

My 2 cents

… would you please elaborate a bit more the scenerio? :wink:

Hi Andy, I use Roundcube. Emails that are marked as spam come from a zabbix5 server in my network. It is highly probable by looking at the logs on rspamd that these emails exceed the SPAM scores or better the thresholds. I know that in the neth7 server it is possible to insert accept rules from userr@domain.xx. What I don’t understand is that if you preniod the spam email and mark it as no spam, the system ignores it.

Hi Michael, I just described the scenario to Andy. Thank you

1 Like

Why zabbix use a .lan account instead of a “real” email address? :wink: Or an Alias one?

Anyway, Zabbix should use a “working SMTP set”, but also should be identifed as “real” email address. You don’t need a mailbox, just an address, even one that “does not exists” like alias, who could allow you to even route/rule the incoming messages.


Hi, it’s not an alias nor a real account, but I gave the server some directives to accept emails without checking the domain of origin. Also to accept them I have inserted the acceptance rule in the filters, so it works. My initial question is that if I move a message that has ended up in spam and I mark it as no spam why is this rule not put into practice? I use roubdcube and Sogo, but moving emails I did it only from roundcube.

@Francesco IMVHO these are only workarounds. You can also tell rSpamd to whitelist your “fake email addresses” (the proper nethserver way is reported here) but it will still be a workaround.
A proper SMTP credentials set will override all the issues related to the message, a “real/plausible” email address (even not a mailbox) will avoid the use of these whitelisting as a loophole to deliver fake messages… :wink:

1 Like

yes correct, but the fact remains that if the same server that invites emails with the same address is cataloged as spam, I wonder whether to move the message from spam to no spam what is the use? I tried to load the message header in HAM but nothing at the moment. my server does not work on port 25 to receive emails from other domains, so to avoid users and passwords I have entered the rules for accepting invalid messages on postfix

Did you already train rspamd with enough spam/ham mails?

Hi Markuz,
practically it was one of the first operations I performed. However even if you load in HAM, as written before it doesn’t work. I believe that the process that analyzes the emails is the usual spamassasin and therefore I would not want to be wrong but the fact that it has been loaded in HAM does not mean Whitelist, but learning over time … I think …

Usually you should see symbols like this when Bayes Filter is working.



But Bayes filter does not filter mail addresses, it’s only about the content.


Yes, in fact, I agree with you. Years ago I used smeserver and I knew that spamassasin filters the content, so much so that as an auxiliary support we used the management of whites and blacklists from repositories. However it is clear that if I have to pass the emails I will do it through the sender’s whitelist, but if I wanted to manage it only from spamasssasin I think it is appropriate to train the HAM filter. Thanks anyway for the reply.

1 Like