Timers determine session liveness and retry behaviour. Graceful Restart and BFD affect failure detection and convergence.
Learning objectives
- Explain Keepalive, Hold and ConnectRetry.
- Understand Hold Time negotiation.
- Explain Graceful Restart and helper behaviour.
- Understand BFD integration and convergence measurement.
- Design timer and recovery QA tests.
1. Core timers
| Timer | Purpose | Notes |
|---|---|---|
| Keepalive | Periodic liveness message | Cisco defaults may be 60 seconds; platform/configuration dependent. |
| Hold Time | Maximum interval without acceptable BGP message | Lower non-zero proposal is used; if either proposes zero, negotiated Hold Time is zero. |
| ConnectRetry | Delay before retrying TCP establishment | Implementation/configuration dependent. |
| Graceful Restart timers | Control stale-route retention and restart/helper behaviour | Capability and platform dependent. |
2. Configuration
router bgp 65001
timers bgp 30 90
neighbor 11.1.1.2 timers 20 60
Verify syntax and precedence on the target release. Avoid changing timers without considering peer compatibility and operational impact.
3. Graceful Restart
- A restarting speaker signals restart capability and relevant state.
- A capable helper may retain eligible stale routes temporarily.
- Routes are refreshed as the restarting speaker re-establishes sessions.
- Stale routes are removed when not refreshed within the applicable period.
Graceful Restart is not a guarantee of uninterrupted forwarding. Success depends on forwarding-plane continuity, capability negotiation, timers, topology and implementation.
4. BFD and BGP
Bidirectional Forwarding Detection provides fast failure detection. BGP can use BFD state to trigger neighbour failure handling faster than a long Hold Timer.
interface GigabitEthernet0/0
bfd interval 300 min_rx 300 multiplier 3
!
router bgp 65001
neighbor 11.1.1.2 fall-over bfd
Illustrative syntax only. Detection depends on negotiated intervals, multiplier, scheduling and platform implementation.
5. Convergence measurement
- Record baseline session, route count and forwarding path.
- Trigger a controlled link or peer failure.
- Measure detection and session-down time.
- Measure route withdrawal or alternate-path selection.
- Measure forwarding interruption and packet loss separately.
- Restore link and record session and route recovery.
- Repeat with timer/BFD/Graceful Restart settings.
Senior QA scenarios
| ID | Test | Expected result |
|---|---|---|
| K-01 | Different Hold proposals | Lower non-zero value; zero proposal yields zero. |
| K-02 | No acceptable BGP message | Hold Timer expires at negotiated limit. |
| K-03 | Change Keepalive | Messages follow configured behaviour. |
| K-04 | Peer TCP unreachable | Retry follows implementation. |
| K-05 | Graceful Restart peer restart | Eligible stale routes retained for applicable period. |
| K-06 | Route not refreshed | Stale route removed after timer. |
| K-07 | BFD failure | BGP reacts to BFD down as supported. |
| K-08 | Vary BFD parameters | Detection reflects negotiated values. |
| K-09 | Restore peer | Session/routes recover. |
| K-10 | Control-plane restart | Data-plane impact recorded separately. |
Interview questions
- Keepalive versus Hold Timer?
- How is Hold Time negotiated?
- What if Hold Time is zero?
- What is ConnectRetry?
- What is Graceful Restart?
- Restarting speaker versus helper?
- Does Graceful Restart guarantee hitless forwarding?
- What is BFD?
- What determines BFD detection time?
- How measure BGP convergence?
Student Knowledge Assessment – Module K (reconstructed)
Q1. What does the BGP Hold Timer measure?
A. Time since last acceptable BGP message
B. TCP retransmission count
C. Route age
D. IGP SPF interval
Correct answer: A. Time since last acceptable BGP message
Explanation: It expires if no acceptable BGP message is received within the negotiated interval.
Q2. How is a non-zero Hold Time generally negotiated?
A. Higher
B. Lower
C. Average
D. Random
Correct answer: B. Lower
Explanation: The lower non-zero proposed value is used.
Q3. What if either peer proposes Hold Time zero?
A. 180 seconds
B. 90 seconds
C. Negotiated Hold Time is zero
D. Session cannot establish
Correct answer: C. Negotiated Hold Time is zero
Explanation: A zero proposal results in a negotiated Hold Time of zero.
Q4. What is ConnectRetry used for?
A. Retrying TCP establishment
B. Refreshing ROAs
C. Selecting MED
D. Reflecting routes
Correct answer: A. Retrying TCP establishment
Explanation: It controls retry timing for TCP session establishment.
Q5. What is Graceful Restart intended to reduce?
A. Forwarding disruption during eligible control-plane restart
B. Prefix-list length
C. AS_PATH length
D. TCP port usage
Correct answer: A. Forwarding disruption during eligible control-plane restart
Explanation: It can preserve eligible stale routes temporarily during restart.
Q6. Does Graceful Restart guarantee no packet loss?
A. Yes
B. No
C. Only eBGP
D. Only with MED
Correct answer: B. No
Explanation: Forwarding continuity depends on platform and topology.
Q7. What does BFD provide?
A. Fast failure detection
B. Route aggregation
C. RPKI signing
D. AS translation
Correct answer: A. Fast failure detection
Explanation: BFD provides rapid detection of forwarding-path failure.
Q8. What influences BFD detection time?
A. Negotiated intervals/multiplier and scheduling
B. Hostname
C. Community name
D. Prefix length
Correct answer: A. Negotiated intervals/multiplier and scheduling
Explanation: Detection depends on negotiated parameters and implementation.
Q9. What should QA measure during failover?
A. Only BGP state
B. Only one ping
C. Detection, route convergence and forwarding impact separately
D. Only uptime
Correct answer: C. Detection, route convergence and forwarding impact separately
Explanation: Control-plane state and forwarding impact are distinct.
Q10. What happens if a stale route is not refreshed in the applicable GR period?
A. Retained forever
B. Removed
C. Becomes static
D. Changes origin
Correct answer: B. Removed
Explanation: Stale routes are removed when not refreshed within the applicable period.