[j-nsp] M7i/M10i - 8.5R4.3 - cfeb RDP: Keepalive timeout for rdp.(scb:39937)
nkhambal at juniper.net
Fri Oct 9 14:11:20 EDT 2009
What is the JUNOS version on this router?
RDP is a Reliable Delivery Protocol. Its an internal TCP-like protocol used
between RE and PFE (CFEB board in this case) to communicate with each and
exchange information such as route, stats, interface status etc. This
communication happens over a socket which connects RE and PFE. Periodic
keepalives are exchanged by RE and PFE to check the health of this
If you are seeing RDP keepalives timeouts and a subsequent CFEB resets, you
should start with checking for the following things.
- is RE CPU running high? (show chassis routing-engine and show system
- is CFEB CPU running high? (show chassis cfeb)
- is CFEB running low on heap/buffer memory? (show chassis cfeb)
- is RE running low on system buffers? ("show system virtual-memory" and
"show system buffers")
- Any type of DDoS attack (ARP storm for example) targeted to the router
interfaces sending too many packets to RE. Do you have a loopback filter
- If sampling is enabled, any high rate of sampling? Sudden increase in
traffic volume may cause more than normal data to be sampled.
- investigate possible configuration and routing policy changes in your
network that would cause this router to attract more traffic than it
earlier did before the problem surfaced.
In Addition gather these commands outputs as well (take at least 2-3
snapshots and 5 mins interval).
- show pfe statistics traffic
- show pfe statistics notification
- show pfe statistics discard
- show pfe statistics error
>From CFEB shell (to log into CFEB, enter shell as root and do "vty cfeb0"
- show packet
- show packet statistics
- show ttp statistics
- show heap 0
- show heap 1
- show route summary
- show syslog messages
- show syslog info
- show nvram
Gather all the information and provide this data to the JTAC case owner.
On 10/9/09 9:57 AM, "Gibson, Aaron F" <aaron.gibson at verizonbusiness.com>
> We having been losing CFEBs like a plague all with the above error (or
> similar) in the logs. No one at Juniper seems to know what RDP is (nearly
> 30 JTAC tickets opened in the last few months) does anyone on this list have
> any insight?
More information about the juniper-nsp