TCP Keep-Alive Mechanism Explained

TCP Keep-Alive is a mechanism to detect whether a connection is alive, used to timely discover and clean up invalid connections.

Keep-Alive Mechanism Principle

Workflow

1. Idle Wait: After the connection is idle for a period of time (default 2 hours), start sending Keep-Alive probe packets
2. Send Probe: Send a TCP segment without data, sequence number is current sequence number minus 1
3. Wait for Response:
   • Receive ACK: Connection is normal, reset timer
   • Receive RST: Connection is reset by the other party, close connection
   • Timeout no response: Connection may be invalid, continue probing

Probe Parameters

• tcp_keepalive_time: How long to wait after connection is idle before sending probes (default 7200 seconds)
• tcp_keepalive_intvl: Interval between probe packets (default 75 seconds)
• tcp_keepalive_probes: Maximum number of probes (default 9 times)

Purpose of Keep-Alive

1. Detect Invalid Connections

• Network interruption, device failure, etc. cause connection failure
• Timely discover and clean up invalid connections, release resources

2. Prevent Connection False Death

• Application layer did not close connection normally, but connection is invalid
• Keep-Alive can detect this situation

3. Keep Connection Active

• Some intermediate devices (such as NAT) will clean up long-idle connections
• Keep-Alive can keep connection from being cleaned up

Disadvantages of Keep-Alive

1. Resource Consumption

• Need to maintain connection state and timers
• With large number of connections, Keep-Alive increases system burden

2. Risk of Misjudgment

• Network latency may cause misjudgment of connection failure
• May mistakenly close normal but high-latency connections

3. Not Timely

• Default wait time is long (2 hours)
• Cannot quickly detect connection failure

Application Layer Heartbeat vs Keep-Alive

Keep-Alive

• Advantage: Operating system level support, no need for application layer implementation
• Disadvantage: Fixed parameters, not flexible enough, long wait time

Application Layer Heartbeat

• Advantage: Flexible and controllable, can carry business data, more timely detection
• Disadvantage: Need application layer implementation, increase development cost

Configuration Example

Linux System Configuration

bash
# View Keep-Alive parameters
sysctl net.ipv4.tcp_keepalive_time
sysctl net.ipv4.tcp_keepalive_intvl
sysctl net.ipv4.tcp_keepalive_probes

# Modify Keep-Alive parameters
sysctl -w net.ipv4.tcp_keepalive_time=600
sysctl -w net.ipv4.tcp_keepalive_intvl=30
sysctl -w net.ipv4.tcp_keepalive_probes=3

Programming Configuration (Python)

python
import socket

sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPIDLE, 600)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPINTVL, 30)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPCNT, 3)

Related Questions

• What is the difference between Keep-Alive and heartbeat packets?
• How to optimize Keep-Alive parameters?
• Why is the default Keep-Alive time so long?