Hello Jon,
Don't worry about my name - I've been renamed from Peter to Paul countless times (though I wonder why...)
Anyway, this is interesting that it made it work. As you probably know, when a TCP session is established, both parties negotiate the maximum size of a TCP segment that can be sent to a particular peer in this session. Usually, this size (abbreviated as MSS) is 1460 bytes, following the fact that with additional TCP header (20B) and IP header (20B), the total IP packet size is 1500 bytes - the usual size.
What the ip tcp adjust-mss command does is actually rewrite the MSS size advertised during TCP connection establishement to a lower value that is specified by this command. In essence, the router affects any TCP session passing through it in a way that it forces the end hosts to send smaller segments. The TCP endpoints do not really know this has happened - to them, it will simply seem that the other party has announced a smaller MSS.
The question is... why did this help in your case? It seems that this service is not purely PPPoA. Now, our goal is to find out the maximum MSS size we can use without problems. This value works but by allowing larger segments, we decrease the overhead.
We will do this by sending a series of pings of increasing size to a responding IP address, say, 8.8.8.8, in which we prohibit any fragmentation. Proceed according to the following transcript:
Router# ping
Protocol [ip]:
Target IP address: 8.8.8.8
Repeat count [5]: 1
Datagram size [100]:
Timeout in seconds [2]:
Extended commands [n]: y
Source address or interface:
Type of service [0]:
Set DF bit in IP header? [no]: yes
Validate reply data? [no]:
Data pattern [0xABCD]:
Loose, Strict, Record, Timestamp, Verbose[none]: v
Loose, Strict, Record, Timestamp, Verbose[V]:
Sweep range of sizes [n]: y
Sweep min size [36]: 1400
Sweep max size [18024]: 1500
Sweep interval [1]:
Type escape sequence to abort.
Sending 101, [1400..1500]-byte ICMP Echos to 8.8.8.8, timeout is 2 seconds:
Packet sent with the DF bit set
Request 0 timed out (size 1400)
Reply to request 1 (243 ms) (size 1401)
Reply to request 2 (159 ms) (size 1402)
Reply to request 3 (227 ms) (size 1403)
Reply to request 4 (461 ms) (size 1404)
Reply to request 5 (344 ms) (size 1405)
Reply to request 6 (235 ms) (size 1406)
Reply to request 7 (201 ms) (size 1407)
Reply to request 8 (67 ms) (size 1408)
Reply to request 9 (58 ms) (size 1409)
Reply to request 10 (51 ms) (size 1410)
Reply to request 11 (50 ms) (size 1411)
Reply to request 12 (50 ms) (size 1412)
Reply to request 13 (51 ms) (size 1413)
Reply to request 14 (67 ms) (size 1414)
Reply to request 15 (50 ms) (size 1415)
Reply to request 16 (50 ms) (size 1416)
Reply to request 17 (50 ms) (size 1417)
Reply to request 18 (51 ms) (size 1418)
Reply to request 19 (58 ms) (size 1419)
Reply to request 20 (51 ms) (size 1420)
Reply to request 21 (42 ms) (size 1421)
Reply to request 22 (51 ms) (size 1422)
Reply to request 23 (50 ms) (size 1423)
Reply to request 24 (50 ms) (size 1424)
Reply to request 25 (42 ms) (size 1425)
Reply to request 26 (84 ms) (size 1426)
Reply to request 27 (59 ms) (size 1427)
Reply to request 28 (50 ms) (size 1428)
Reply to request 29 (51 ms) (size 1429)
Reply to request 30 (50 ms) (size 1430)
Reply to request 31 (59 ms) (size 1431)
Reply to request 32 (50 ms) (size 1432)
Reply to request 33 (50 ms) (size 1433)
Reply to request 34 (59 ms) (size 1434)
Reply to request 35 (51 ms) (size 1435)
Reply to request 36 (50 ms) (size 1436)
Reply to request 37 (50 ms) (size 1437)
Reply to request 38 (59 ms) (size 1438)
Reply to request 39 (51 ms) (size 1439)
Reply to request 40 (58 ms) (size 1440)
Reply to request 41 (51 ms) (size 1441)
Reply to request 42 (58 ms) (size 1442)
Reply to request 43 (51 ms) (size 1443)
Reply to request 44 (50 ms) (size 1444)
Reply to request 45 (50 ms) (size 1445)
Reply to request 46 (59 ms) (size 1446)
Reply to request 47 (50 ms) (size 1447)
Reply to request 48 (51 ms) (size 1448)
Reply to request 49 (50 ms) (size 1449)
Reply to request 50 (59 ms) (size 1450)
Reply to request 51 (50 ms) (size 1451)
Reply to request 52 (50 ms) (size 1452)
Reply to request 53 (59 ms) (size 1453)
Reply to request 54 (50 ms) (size 1454)
Reply to request 55 (50 ms) (size 1455)
Reply to request 56 (50 ms) (size 1456)
Reply to request 57 (59 ms) (size 1457)
Reply to request 58 (50 ms) (size 1458)
Reply to request 59 (50 ms) (size 1459)
Reply to request 60 (59 ms) (size 1460)
Reply to request 61 (50 ms) (size 1461)
Reply to request 62 (51 ms) (size 1462)
Reply to request 63 (159 ms) (size 1463)
Reply to request 64 (50 ms) (size 1464)
Reply to request 65 (59 ms) (size 1465)
Reply to request 66 (50 ms) (size 1466)
Reply to request 67 (50 ms) (size 1467)
Reply to request 68 (59 ms) (size 1468)
Reply to request 69 (51 ms) (size 1469)
Reply to request 70 (50 ms) (size 1470)
Reply to request 71 (126 ms) (size 1471)
Reply to request 72 (58 ms) (size 1472)
Reply to request 73 (51 ms) (size 1473)
Reply to request 74 (50 ms) (size 1474)
Reply to request 75 (59 ms) (size 1475)
Reply to request 76 (59 ms) (size 1476)
Reply to request 77 (50 ms) (size 1477)
Reply to request 78 (50 ms) (size 1478)
Reply to request 79 (50 ms) (size 1479)
Reply to request 80 (50 ms) (size 1480)
Reply to request 81 (50 ms) (size 1481)
Reply to request 82 (50 ms) (size 1482)
Reply to request 83 (51 ms) (size 1483)
Reply to request 84 (50 ms) (size 1484)
Reply to request 85 (51 ms) (size 1485)
Reply to request 86 (50 ms) (size 1486)
Reply to request 87 (59 ms) (size 1487)
Reply to request 88 (50 ms) (size 1488)
Reply to request 89 (50 ms) (size 1489)
Reply to request 90 (51 ms) (size 1490)
Reply to request 91 (50 ms) (size 1491)
Reply to request 92 (51 ms) (size 1492)
Request 93 timed out (size 1493)
Request 94 timed out (size 1494)
Request 95 timed out (size 1495)
Request 96 timed out (size 1496)
Request 97 timed out (size 1497)
Request 98 timed out (size 1498)
Request 99 timed out (size 1499)
Request 100 timed out (size 1500)
Success rate is 91 percent (92/101), round-trip min/avg/max = 42/71/461 ms
Note that here, we start losing packets with the total packet size of 1493, with 1492 bytes being the last usable size. The corresponding TCP MSS would therefore be 40 bytes less, i.e. 1452. Following this experiment, the Dialer0 would be configured using two additional commands:
ip mtu 1492
ip tcp adjust-mss 1452
Would you mind performing a similar experiment? Before running it, though, remove any ip mtu or mtu commands from your Dialer0 interface.
Best regards,
Peter
