Problem
Sendmail fails to send or deliver mail and popper fails to retrieve mail for POP clients. SMTP connections to port 25 are refused. The file /usr/adm/syslog shows:
sendmail: rejecting connection on port 25 load average ##
where ## is the system's load average limit (usually 18). The
process table shows many (usually over 10) sendmail or popper
processes. The mail queue fills up, and sar shows a very low
percent idle, and a high percent system usage. Cause
When the load average on a server (the average number of jobs in the run queue in the last minute) gets too high, sendmail will compensate by refusing to accept SMTP connections on port 25. The excessive load average can be caused by a system being overly busy or it can be caused by poorly designed or misconfigured software. Excessive load averages on POP and SMTP servers are usually caused by one of the following:
1. Users are trying to pop mail for a nonexistent account (or an account without a mailbox) on the server.
2. Popper processes cannot keep up with mailbox parsing due to users with extremely large mailboxes setting their mail programs (such as Netscape, Eudora, and so on) to check for new mail every 1 minute (or some short amount of time). Large mailboxes are usually the result of users configuring their mail programs to leave mail on the server after popping it, then neglecting to clear the server mailbox periodically.
3. Mailbox ownership and/or permissions are incorrect.
4. Lock files in /tmp are not being released.
5. The default timeout parameter for popper is too high for the system load.
6. The RefuseLA (X) option in the sendmail configuration file has been set to a low value.
Solution
Follow these steps:1. Make sure all accounts on the server have a mailbox and that its permissions and ownership are correct. Mailboxes should be owned by the user and belong to group sys. Permissions should be -rw-------.
2. Check the process table for an overload of popper or sendmail processes:
# ps -ef | grep popper
# ps -ef | grep sendmail3. If there are several (more than 10) popper processes running, you will need to change the timeout (-T) option for popper with the following steps:
Edit the /etc/inetd.conf file and add the -T 15 option to the /etc/popper line. The edited line should look like:
pop3 stream tcp nowait root /etc/popper -T 15 popperSave and exit the file, then restart inetd with:
# kill -1 `cat /etc/inetd.pid`Note: The characters surrounding the cat command are not single quotes. They are back-tics (usually located on the ~ key).
4. Check for lock files in /tmp that are not being released. Lock files are of the format ####.### and contain the PID of the popper process that is locking the mailbox. Remove any that are older than 5 minutes.
5. Check the /usr/lib/sendmail.cf file for the value of the RefuseLA option (OX). If it is not listed in the file, then sendmail is using the default value of 12, which should be fine. If the value has been changed to a value below 10, reset it to the default value of 12, save and exit the sendmail.cf file, and restart the sendmail daemon with:
# kill -1 `cat /etc/sendmail.pid`Note: The characters surrounding the cat command are not single quotes. They are back-tics (usually located on the ~ key).
6. If none of the above steps addresses the problem, edit the
/etc/inetd.conf file and add -t and -d to the /etc/popper
command to get popper's debug information written to the file
/usr/adm/syslog, and trace information written to the file
/tmp/pop.out. The edited line in /etc/inetd.conf would look like:pop3 stream tcp nowait root /etc/popper -t /tmp/pop.out -d popperAfter making these changes, restart the inetd daemon with the kill command listed in step 3.
See also
ta105554, "Tips for troubleshooting POP mail."ta110212, "Some clients attempting to pop mail from the server are unsuccessful."
ta107717, "Clients popping mail from the server are causing files to be left in /tmp on OSR5."