MTU Change Generated Large Volume of Traffic and Caused SA6500 Cluster to Fail

[dfry_]

Greetings All,

We have a 2-node SA6500 cluster that was having problems communicating with Terminal Servers over a fast (OC12) WAN link. We changed the MTU on the internal Juniper interfaces from 1500 to 1380 and this appeared to correct the issue, however a couple of hours later we started to have major problems with about 50 users on the system. We noted heavy (200+Mbit) traffic on the internal interfaces and around 20% CPU utilization, both far higher than we had seen before. The cluster went up and down and some users got disconnected. When the MTU was changed back to 1500 everything returned to normal fairly quickly. What did we do wrong? Has anyone run into similar problems? Please advise.

Thanks .. Dan

[Mrkool_]

any chance you did a tcp dump on the internal interface while the problem was happening?

[dfry_]

It turns out that the MTU size option DOES NOT WORK in 6.5r2 and causes the system to fail if it is changed.

Thanks,

Dan