Redmine

I will add:

On a working server, the output from /var/log/maillog is:

Dec 6 14:12:53 helpdesk postfix/smtpd[3655]: connect from localhost.localdomain[127.0.0.1]
Dec 6 14:12:53 helpdesk postfix/smtpd[3655]: 4F46048682: client=localhost.localdomain[127.0.0.1]
Dec 6 14:12:53 helpdesk postfix/cleanup[3658]: 4F46048682: message-id=<redmine.journal-376785.20231206141252@mydomain.cz>
Dec 6 14:12:53 helpdesk postfix/qmgr[2630]: 4F46048682: from=<helpdesk@mydomain.cz>, size=3233, nrcpt=17 (queue active)
Dec 6 14:12:53 helpdesk postfix/smtpd[3655]: disconnect from localhost.localdomain[127.0.0.1]

...on a broken server:

Dec 6 14:13:06 helpdesk-clone postfix/smtpd[2889]: connect from localhost.localdomain[127.0.0.1]
Dec 6 14:13:06 helpdesk-clone postfix/smtpd[2889]: warning: Illegal address syntax from localhost.localdomain[127.0.0.1] in RCPT command: <>
Dec 6 14:13:06 helpdesk-clone postfix/smtpd[2889]: disconnect from localhost.localdomain[127.0.0.1]

I would check if the version of postfix has changed:

$ postconf -d | grep mail_version
mail_version = 3.5.6
milter_macro_v = $mail_name $mail_version

as a new version could have become stricter regarding the SMTP protocol.

If the version is the same in both servers, I would compare the RCPT in the SMTP messages sent by redmine to the server with ngrep:

$ ngrep -d any -q -W byline RCPT 

The servers are very very old.

redmine version: 1.2.1
OS: CentOS 5

The ngrep tool is missing and won't install. It has more dependencies and CentOS is no longer supported.
I have the migration ticket here: https://www.redmine.org/boards/2/topics/69117 ... but I haven't received much advice yet :-)
The versions are the same.

A server where emails don't work:
[root@clone ~]# postconf -d | grep mail_version
mail_version = 2.3.3
milter_macro_v = $mail_name $mail_version

Server where emails works
[root@restore ~]# postconf -d | grep mail_version
mail_version = 2.3.3
milter_macro_v = $mail_name $mail_version

Ladislav Kudláček wrote in RE: Email notifications have stopped being sent:

The servers are very very old.

redmine version: 1.2.1
OS: CentOS 5

The ngrep tool is missing and won't install. It has more dependencies and CentOS is no longer supported.
After solving this problem, we will migrate the data to the new operating system and the new redmine.
I have the migration ticket here: https://www.redmine.org/boards/2/topics/69117 ... but I haven't received much advice yet :-)

The versions postfix are the same.

A server where emails don't work:
[root@clone ~]# postconf -d | grep mail_version
mail_version = 2.3.3
milter_macro_v = $mail_name $mail_version

Server where emails works
[root@restore ~]# postconf -d | grep mail_version
mail_version = 2.3.3
milter_macro_v = $mail_name $mail_version

We tested:
On the server where e-mails are not being sent, we restored the database from the server where e-mails were still being sent (as of 1/11/2023). The result was that the mails started to be sent. As soon as we restored the database to this same server with current data, the mails stopped sending again...I think the problem will be somewhere in the database. Any ticket with bad ASCII encoding? I have no idea yet...

1. echo "test" | mail -s "test" myname@mydomain.cz
   Dec 7 02:17:42 helpdesk-clone postfix/pickup¹⁷³¹³: D2B6D48730: uid=0 from=<root>
   Dec 7 02:17:42 helpdesk-clone postfix/cleanup¹⁷⁷⁶⁷: D2B6D48730: message-id=<20231207011742.D2B6D48730@helpdesk-clone.mydomain.local>
   Dec 7 02:17:42 helpdesk-clone postfix/qmgr²⁵⁵²: D2B6D48730: from=<root@helpdesk-clone.mydomain.local>, size=333, nrcpt=1 (queue active)
   Dec 7 02:18:12 helpdesk-clone postfix/smtp¹⁷⁷⁶⁹: connect to gw.mydomain.cz[172.16.0.6]: Connection timed out (port 25)
   Dec 7 02:18:12 helpdesk-clone postfix/smtp¹⁷⁷⁶⁹: D2B6D48730: to=<myname@mydomain.cz>, relay=none, delay=30, delays=0.02/0.01/30/0, dsn=4.4.1, status=deferred (connect to gw.mydomain.cz[172.16.0.6]: Connection timed out)

Postfix works. The problem will be somewhere in redmine system (smtp port is blocked by firewall, "Connection timed out" is correct)

I would try to do a diff between the old and new server relevent folders like this:

diff -r old_redmine new_redmine

If this is generate a too large diff output, I would check just the config folder:

diff -r old_redmine/config/ new_redmine/config/

If nothing relevant seems to differ in redmine, I would also try to compare the postfix configuration:

diff -r /etc/postfix postfix_folder_from_the_other_server

Obs: do not post the results here unless you know there are no confidential information.

Regarding ngrep, you could capture the data using tcpdump:

tcpdump -i any -s 0 -w test.cap 'port 25'

I made exactly these distinctions. There really was a difference in the config, I adjusted it according to the working server, but it didn't help. I always edit confidential information (myname, mydomain etc...). So the output is modified. There is no difference in postfix, it is set identically... and according to the test from the command line, mail is sent via postfix. It doesn't just send from redmine.

Ugh....resolved. The problem was in the mailbox of two users. There were hundreds of users. By method of elimination, I found that two users were responsible. Someone set mailbox usernames with accents.

Thanks very much Mayama Takeshi for reaction...