Brad Zobrist wrote:
I am using forkserver, the server is not thrashing (it's a 15k scsi box and has avg TPS of ~250),
I was speaking of resource starvation due to not enough RAM or too many processes, not disk-thrashing. What does `top` report for the load? How much RAM is installed and how big is your swap partition (especially if you installed the more RAM after you created the swap partition)?
I have resolved the issue. The issue was with a BIG-IP that is in the picture.
If you don't have multiple devices behind the load-balancer servicing SMTP, there is *no* reason to use the BIG-IP here at all. I would strongly recommend moving the SMTP server out from behind the load-balancer for testing purposes and probably for production, too.
I do have another question regarding timeouts. I'm seeing a ~5 second delay to get the 220 response. Where is this defined? Is it customizable?
Load check_earlytalker with a parameter of 1 and see if it makes a difference. It is possible that your system load is so high (or your RAM is so low) that it is taking ~5 seconds to fork a new process. You may want to try the pre-fork server (but only after someone has applied the pending patches to it).
John -- John Peacock Director of Information Research and Technology Rowman & Littlefield Publishing Group 4501 Forbes Boulevard Suite H Lanham, MD 20706 301-459-3366 x.5010 fax 301-429-5748
