Applicable to:
- Plesk Onyx for Linux
Symptoms
Mail delivery does not work periodically. The following messages can be found in /var/log/maillog:
CONFIG_TEXT: hostname postfix/smtpd[2177845]: warning: read TCP map reply from 127.0.0.1:12346: unexpected EOF (Connection timed out)
hostname postfix/smtpd[2177845]: warning: read TCP map reply from 127.0.0.1:12346: unexpected EOF (Connection timed out)
hostname postfix/smtpd[2177845]: warning: tcp:127.0.0.1:12346 lookup error for "john.doe1@example.com"
hostname postfix/smtpd[2177845]: NOQUEUE: reject: RCPT from mailexample.com[203.0.113.2]: 451 4.3.0 <john.doe1@example.com>: Temporary lookup failure; from=<john.doe2@example2.com> to=<john.doe1@example.com> proto=ESMTP helo=<mail.example.com>
hostname postfix/smtpd[2177845]: lost connection after DATA from mail.example.com[203.0.113.2]
hostname postfix/smtpd[2177845]: disconnect from mail.example.com[203.0.113.2]
Cause
Not clear. Temporary Postfix or SRS (Sender Rewriting Scheme) issue.
Resolution
-
Connect to the server via SSH
-
Check the
postfix-srsline in/etc/postfix/master.cfconfiguration file. The correct output should be as follows:# grep postfix-srs /etc/postfix/master.cf
127.0.0.1:12346 inet n n n - - spawn user=popuser:popuser argv=/usr/lib/plesk-9.0/postfix-srs -
Verify that
/proc/sys/net/ipv4/tcp_fin_timeoutis equal or lower than 60 seconds:# cat /proc/sys/net/ipv4/tcp_fin_timeout
30 -
Set Maximum number of connections for a user per IP address to default value
10going to Plesk > Tools & Settings > Mail Server Settings -
If the steps above do not help or the configuration is correct, try to temporary disable and then re-enable SRS back according to the following article: How to turn off Sender Rewriting Scheme support?
Comments
9 comments
My server just started getting this error.
Restarting the service seems to get it stable for 5 minutes or so.
The output matches the above. Adding the -s as stated in the turning off SRS thread has no impact on whether it occurs or not.
Any other suggestions?
@John
Does your server use CentOS? If so, you may also check SELinux (https://support.plesk.com/hc/en-us/articles/360001677533 or https://support.plesk.com/hc/en-us/articles/360003479954-Unable-to-send-and-email-in-Plesk-on-server-with-SELinux-enabled-Unable-to-open-file-var-lib-plesk-mail-srs-secret-Permission-denied)
Yes CentOS , SElinux is diabled
@John
Then it is needed to investigate the issue thoroughly. Please submit a support request using your available option.
I Have the same problem.
Configuration OK
grep postfix-srs /etc/postfix/master.cf
127.0.0.1:12346 inet n n n - - spawn user=popuser:popuser argv=/usr/lib64/plesk-9.0/postfix-srs
Selinux disabled
sestatus
SELinux status: disabled
Maillog
Nov 20 15:48:42 s3 postfix/smtpd[19836]: warning: read TCP map reply from 127.0.0.1:12346: unexpected EOF (Connection timed out)
Nov 20 15:48:42 s3 postfix/smtpd[19836]: warning: tcp:127.0.0.1:12346 lookup error for "mymail@mydomain.com"
Hi @Robert,
Are there any messages like the one below in the maillog as well?
warning: /usr/lib64/plesk-9.0/postfix-srs: process id #####: command time limit exceeded
Yes! a lot
Nov 26 03:56:53 s3 postfix/spawn[5033]: warning: /usr/lib64/plesk-9.0/postfix-srs: process id 5043: command time limit exceeded
Nov 26 03:56:53 s3 postfix/spawn[5052]: warning: /usr/lib64/plesk-9.0/postfix-srs: process id 5054: command time limit exceeded
Nov 26 03:56:53 s3 postfix/spawn[5056]: warning: /usr/lib64/plesk-9.0/postfix-srs: process id 5057: command time limit exceeded
Then let me provide you with some tips on how to troubleshoot this issue. Basically in can occur in case there is one IP that connects to the server and then does not close the connection for some reason.
1. Find all 'max connection count' in the maillog, it should looks like the example below:
# grep 'max connection count' /var/log/maillog | tail -n 2
postfix/anvil[28023]: statistics: max connection count 34 for (smtp:203.0.113.2 ) at Sep 3 11:47:48
postfix/anvil[51714]: statistics: max connection count 11 for (smtp:203.0.113.2 ) at Sep 3 12:25:57
In this example, there are 34 and 11 connections from 203.0.113.2 .
2. Then, check this IP with the netstat in order to understand whether or not there are some hanging connections:
# netstat -tulnap | grep 203.0.113.2 | wc -l
23
# netstat -tulnap | grep 203.0.113.2 | head -n 2
tcp 0 1 203.0.113.10:25 203.0.113.2:49701 FIN_WAIT1 -
tcp 0 1 203.0.113.10:25 203.0.113.2:49784 FIN_WAIT1 -
3. Now it's time to check Connect/Disconnect ratio from offending IP:
This is the normal output (connect matches disconnect):
# grep 'connect .*203.0.113.2' /var/log/maillog | awk '{print $6}' | sort | uniq -c
603 connect
603 disconnect
This is not correct output:
# grep 'connect .*203.0.113.2' /var/log/maillog | awk '{print $6}' | sort | uniq -c
86 connect
20 disconnect
4. In theory, the following script is intended to provide IPs addresses connection ratio statistics:
# for IP in `grep 'max connection count' /var/log/maillog | awk '{ print $12 }' | sort | uniq | sed 's/(smtp://g; s/)//g;'`; do echo -e "Checking IP.. $IP"; egrep "connect .*$IP" /var/log/maillog | awk '{print $6}' | sort | uniq -c ; done
Overall in case on step 3 you are getting connect/disconnect ratio mismatch, this means that host 203.0.113.2 does not close the connection and is causing the issue. Solution would be to either ban this IP address or contact them somehow letting them know that their mail client is malfunctioning.
As a workaround, it should be also possible to increase number of connection from one IP address at Tools & Settings > Mail Server Settings.
Ok thanks, I try to activate fail2ban
Please sign in to leave a comment.