Mail Flow Within a Cluster

benny.villa_ironport · ‎05-02-2007

Quick Question,

I have 4 C600 in a Cluster (4.7.1)

2 of my C600's have over 5000 Messages in its Workqueue and the other 2 have less than 300 each, Anyone have any ideas as to why this could happen within a Cluster Setup ? Any settings i can look at?

Please email me Benny.Villa@us.army.mil Thanks !

kluu_ironport · ‎05-08-2007

Make sure that either your MX records are ordered in a round robin fashion in order to ensure that incoming mail gets an equal chance to hit all of the machines. Below is a link to an article that illustrates this:

http://www.zytrax.com/books/dns/ch9/rr.html

Quick Question,

I have 4 C600 in  a Cluster (4.7.1)

2 of my C600's have over 5000 Messages in its Workqueue and the other 2 have less than 300 each,  Anyone have any ideas as to why this could happen within a Cluster Setup ?  Any settings i can look at?

Please email me   Benny.Villa@us.army.mil  Thanks !

Donald Nash · ‎05-21-2007

We have six appliances handling our inbound mail (3x C60 and 3x C600), but we have a single MX record which ultimately points to an IP address that is managed by a load balancer. This works wonders at preventing uneven loading, and also makes it easier to add/remove appliances from the service as necessary. I highly recommend it.

We have connection persistence turned on, with a sticky time of one hour. This is necessary for SenderBase rate limiting to work effectively, since otherwise a suspicious sender would get connected to a different appliance each time it connects. This makes for occasional unbalanced loads, when a hot sender gets nailed to a particular appliance. However, it happens very seldom (maybe once a month or so), and I've never seen it fail to sort itself out within half an hour or so.