[c-nsp] PPPOE intermittently fails to respond to PPP keepalives on 7204xvr with 12.4(24)T5

[Richard Haakma]

Hi.

I have a 7204VXR (yes I know it is old) which is serving broadband
customers over ethernet. Another party supplies the last mile and we get
a Gigabit ethernet connection with QinQ where end users get a VLAN each.

e.g.

interface GigabitEthernet0/3.2
 encapsulation dot1Q 2 second-dot1q any
 pppoe enable group mypppoe

etc.

and using this

interface Virtual-Template2
 mtu 1492
 ip unnumbered Loopback0
 ip tcp adjust-mss 1452
 peer default ip address pool dsl-pool
 ppp authentication pap dsl-auth
 ppp authorization dsl-authorize
 ppp accounting dsl-account
 ppp ipcp address required

AAA is by radius.

The problem is that intermittently IOS on the 7204VXR will not respond
to PPP keepalive packets. Some CPEs send keepalives and when they don't
get a reply they disconnect the PPP session and reconnect. Some
customers notice the short outage and it can be repeated several times
over a few minutes. This has been verified by observing packets at the
CPE.

I think that PPP keepalives are handled by a CPU process and something
may be stopping the process from running in time. The 7204 is not
heavily loaded and we first saw this behaviour with only a handful of
connections. I attempted to recreate it in a lab setup with one PPPOE
client on a test BRAS but that runs cleanly.

Also possible is that the last mile provider is dropping keepalive or
keepalive replies. They may drop when the QOS bits are set incorrectly
but this is unlikely as the pattern is intermittent.

Please help, has anyone seen this kind of problem and fixed it up? Has
this been changed in 12.4(24)T8 ?


Regards,
Richard Haakma.