Problem with ppp multilink across PRI

[admin_2]

During disaster recovery testing we were using a full PRI connection (30 B channels). We found that when a third line was opened the data stopped going across the connection and the connection timed out. We eventually found that the problem was related to the IOS. The devices at each end were 3620's with a PRI/Fast Ethernet card. We took 3 IOS versions which were : 12.0(4)T, 12.1(2)T, 12.2(6). All were IP Only. The problem only happened using 12.1(2)T. Testing was done by having an FTP server at one end and transferring a large file across to a PC at the other end. For the 12.0 and 12.2 the lines raised until all 30 lines were in use but with the 12.1 as soon as it opened a 3rd channel the connection was dropped. Tested this 3 times from either end to make sure it wasn't the line. The configurations were not changed. Surely a T train IOS shouldn't have a fundamental problem such as this ?? Can anyone else confirm if they have the same problem.

[#2]

What type of traffic and how is it configured? I have seen sequence problems with Multilink.

Do you have the ppp multilink command on the s0:15 interface as well as the dialer? if not, add it.

What can happen, particularly if bridging over the link is that two packets are sent from the source a big one then a little one.

The big one goes down the first channel, then the second one starts down the second channel. That completes at the destination first and can be forwarded ahead of the other packet. Depending on the traffic, this can be a major problem - ISO transport can have very serious problems with out of sequence packets, and ECMA is worse as an example.

The output of a sh ppp mul may be interesting.