I know this has been discussed several times in the forums, but I am still unclear and would appreciate some clarification. I currently have several RB450G's running 5.2 that are acting as PPPoE Servers in the field. We are noticing that occasionally, we will have a user that complains that he is not able to access certain sites, or the sites load very slowly. It appears we have a MTU issue. Only maybe 2 out of 50 users on any given RB450G are experiencing this problem.
After reading the FAQ and the forums, it appears there are a few solutions. One is to set the "Change MSS" option on the PPPoE profile. This isn't the best solution though, since it creates dynamic MSS mangle rules for every PPPoE session which can be taxing on the RB's CPU with so many PPPoE users. Another, is to create one static mangle firewall rule as the FAQ states:
I added this rule to my RB, but the user is still experiencing issues. At the user end of the client connection, my CPE (Motorola Canopy SM) is connected directly to the end user's computer. The SM is performing NAT. The SM and the PPPoE server are reporting a MTU of 1480. This should leave 12bytes of overhead for PPPoE which should be plenty (only need 8bytes).Use /ip firewall mangle to change MSS (maximum segment size) 40 bytes less than your connection MTU. For example, if you have encrypted PPPoE link with MTU=1492, set the mangle rule as follows:
/ ip firewall mangle
add chain=forward protocol=tcp tcp-flags=syn action=change-mss tcp-mss=!0-1448 new-mss=1448
The FAQ states that you should create the MSS rule to be 40 bytes less than your MTU. It then says that if you have a PPPoE link with a MTU of 1492, to change the MSS to 1448. That is 44bytes, not 40bytes as the instructions state. Am I missing something?
Could the MTU settings on the client's PC be causing this problem? How can I go about further troubleshooting this issue?