The customer reported that one of their BGP session didn't come up after increasing prefix-limit.
Neighbor was shut down on Apr 22 17:25:59.052 due to prefix-limit violation:
Apr 22 17:25:59.052 router-name rpd[30378]: BGP_PREFIX_THRESH_EXCEEDED: 10.0.43.7 (External AS 65530): Configured maximum prefix-limit threshold(4) exceeded for inet-unicast nlri: 5 (instance master)
Apr 22 17:25:59.052 router-name rpd[30378]: BGP_CEASE_PREFIX_LIMIT_EXCEEDED: 10.0.43.7 (External AS 65530): Shutting down peer due to exceeding configured maximum prefix-limit(5) for inet-unicast nlri: 6 (instance master)
<snip>
Apr 22 17:44:31.295 router-name rpd[30378]: bgp_pp_recv:4915: NOTIFICATION sent to 10.0.43.7+38427 (proto): code 6 (Cease) subcode 1 (Maximum Number of Prefixes Reached), Reason: no group for 10.0.43.7+38427 (proto) from AS 65530 found (peer idled due to prefix-limit violation) in master(ae12.0), dropping him
prefix-limit maximum value changed - increased from 5 to 20:
Apr 22 17:44:37.371 router-name sshd[42258]: Accepted password for user11_rw from 10.20.0.40 port 52912 ssh2
Apr 22 17:44:47.329 router-name mgd[42262]: UI_DBASE_LOGIN_EVENT: User 'user11_rw' entering configuration mode
Apr 22 17:44:52.067 router-name mgd[42262]: UI_CMDLINE_READ_LINE: User 'user11_rw', command 'delete protocols bgp group test-bgp family inet unicast prefix-limit maximum 5 '
Apr 22 17:44:52.171 router-name mgd[42262]: UI_CMDLINE_READ_LINE: User 'user11_rw', command 'set protocols bgp group test-bgp family inet unicast prefix-limit maximum 20 '
Apr 22 17:45:01.980 router-name mgd[42262]: UI_COMMIT: User 'user11_rw' requested 'commit synchronize' operation (comment: commit_by__user1_from_host_device)
Apr 22 17:45:07.905 router-name mgd[42262]: UI_COMMIT_COMPLETED: : commit complete
Apr 22 17:45:07.932 router-name mgd[42262]: UI_DBASE_LOGOUT_EVENT: User 'user11_rw' exiting configuration mode
Peer stayed down:
Apr 22 17:46:59.341 router-name rpd[30378]: bgp_pp_recv:4915: NOTIFICATION sent to 10.0.43.7+35541 (proto): code 6 (Cease) subcode 1 (Maximum Number of Prefixes Reached), Reason: no group for 10.0.43.7+35541 (proto) from AS 65530 found (peer idled due to prefix-limit violation) in master(ae12.0), dropping him
Apr 22 17:49:27.349 router-name rpd[30378]: bgp_pp_recv:4915: NOTIFICATION sent to 10.0.43.7+40577 (proto): code 6 (Cease) subcode 1 (Maximum Number of Prefixes Reached), Reason: no group for 10.0.43.7+40577 (proto) from AS 65530 found (peer idled due to prefix-limit violation) in master(ae12.0), dropping him
Apr 22 18:08:55.098 router-name rpd[30378]: bgp_pp_recv:4915: NOTIFICATION sent to 10.0.43.7+35525 (proto): code 6 (Cease) subcode 1 (Maximum Number of Prefixes Reached), Reason: no group for 10.0.43.7+35525 (proto) from AS 65530 found (peer idled due to prefix-limit violation) in master(ae12.0), dropping him
Related configs:
Idle timeout configured is 45 minutes:
[...]
group test-bgp {
type external;
description "test-bgp";
import policy-name;
family inet {
unicast {
prefix-limit {
maximum 20; ----- increased from 5
teardown {
90;
idle-timeout 45; >>>>>>>>> After the session is torn down, it will be kept down for 45 minutes before it is reestablished.
}
export [ test-bgpA test-bgp ];
peer-as 65530;
multipath;
neighbor 10.0.43.5 {
description routerc;
neighbor 10.0.43.7 {
description routerd;
user11_rw@router-name# set protocols bgp group test-bgp family inet unicast prefix-limit maximum 20 teardown 90 idle-timeout ?
Possible completions:
<[Enter]> Execute this command
<timeout> Timeout value, in minutes, for restarting peer (1..2400)
forever Idle the peer until the user intervenes
| Pipe through a command
[edit]
Unless you do "clear bgp neighbor <peer-ip>" after increasing prefix-limit the session will stay down for the amount of time configured (45 minutes) then it will try to establish the session.
The session came up on Apr 22 18:10:59.160 - which is 45 minutes after it was shut down (Apr 22 17:25:59.052) for violating prefix-limit.
Apr 22 18:10:59.160 router rpd[30378]: RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.0.43.7 (External AS 65530) changed state from OpenConfirm to Established (event RecvKeepAlive) (instance master)
If the "idle-timeout" statement is not included, the session will try to reestablish in a short time.