Great delay when sending emails

Hi, fellows,

I have a Linux Centos Server, running Apache, Postfix, Dovecot, Postgrey, Amavis, Spamassassin and all the normal stuff. It is the POP / SMTP Server for all of our customers' accounts.

Our customers (and we too) have realized that whenever they sent an email, through any mail client program (Outlook, Eudora, Thunderbird, no matter if it is from a Windows or Linux environment), that initial time to connect, auth, login, is being very very delayed. Sometimes it even goes into time out. And when this happens, normally the CPU times are quite good, no overload.

Could it be any parameter setting at /etc/postfix/main.cf or other System files, that should be more accurate and better tuned?

I am not so skillful in Linux, but I have total access to all the System files at this Server. I thank you in advance for any help or hint.

Mario Lima./

Don't know much about postfix...I'm sure other's will jump in there...but,
To what port are you having your customers connect to send mail?
What tells you/indicates that there is a delay?

Originally Posted by scasey Don't know much about postfix...I'm sure other's will jump in there...but, To what port are you having your customers connect to send mail? What tells you/indicates that there is a delay?

Hi, scasey. They connect to Ports 110 (POP) and 587 (SMTP). I compare this Server to another I have, exactly like this, in the same Hosting service and network. Sometimes, to send a simple email just saying "Hello", no attachments, it delays until 2 minutes.

Originally Posted by Mario Lima Hi, scasey. They connect to Ports 110 (POP) and 587 (SMTP). I compare this Server to another I have, exactly like this, in the same Hosting service and network. Sometimes, to send a simple email just saying "Hello", no attachments, it delays until 2 minutes.

Where is the delay exactly? From when the email goes from Outbox to Sent folder?

Originally Posted by ilesterg Where is the delay exactly? From when the email goes from Outbox to Sent folder?

Hi, ilesterg, yes, exactly.

Do a live test and see what's being logged in /var/log/maillog and how fast things happen. You'd be interested with the time the connection is received by postfix from the client, the time it sent the message to the relay, and so on. That would at least give you a broad idea of where the bottleneck happens.

Hi, ilesterg,

I did it. I use an ancient email program called Eudora, and it shows in a Window the sending process steps.

The first step is always "Connecting to..." the Smtp Server. I was looking at this Server's maillog with tail command. After this it does the AUTH / LOGIN process, and only from this point on the Server's maillog registers something. As a matter of fact the longest delay time is during the "Connecting to" process.

Detail: I have another Server, hosted at this same ISP, and whenever I sent emails through it, I don't have this long delay time.

What could it mean?

Originally Posted by Mario Lima Hi, ilesterg, I did it. I use an ancient email program called Eudora, and it shows in a Window the sending process steps. The first step is always "Connecting to..." the Smtp Server. I was looking at this Server's maillog with tail command. After this it does the AUTH / LOGIN process, and only from this point on the Server's maillog registers something. As a matter of fact the longest delay time is during the "Connecting to" process. Detail: I have another Server, hosted at this same ISP, and whenever I sent emails through it, I don't have this long delay time. What could it mean?

Could be a dns related problem, meaning that the name resolution takes too long.
Use the same nameservers for your mail server (in /etc/resolv.conf), as those of the second server and see if it helps.

Hi, bathory, great, this is a good idea. I will check it with the Provider.

bathory, we altered using for both the Google Servers Name, 8.8.8.8 and 8.8.4.4. It seems we had an improvement. I'll be sure about his on next week. Thanks a lot.