When establishing SSH connections between Ubuntu clients and CentOS servers, many developers encounter persistent connection drops during idle periods, culminating in the frustrating "Write failed: Broken pipe" error. This behavior typically indicates TCP-level connection timeouts rather than an SSH protocol failure.

# Check current SSH server keepalive settings
grep -E "ClientAlive|TCPKeepAlive" /etc/ssh/sshd_config

# Typical output showing default (often problematic) settings:
# TCPKeepAlive yes
# ClientAliveInterval 0
# ClientAliveCountMax 3

The critical parameters needing adjustment are:

• ClientAliveInterval: Time (seconds) between keepalive packets (recommended: 60-120)
• ClientAliveCountMax: Number of failed keepalives before termination (recommended: 3-5)

# Edit the SSH daemon configuration
sudo nano /etc/ssh/sshd_config

# Add or modify these lines:
ClientAliveInterval 60
ClientAliveCountMax 5
TCPKeepAlive yes

After saving changes, restart the SSH service:

sudo service sshd restart  # CentOS 5.5 syntax
# For modern systems:
# sudo systemctl restart sshd

For additional reliability, configure your SSH client (~/.ssh/config):

Host *
    ServerAliveInterval 30
    ServerAliveCountMax 5
    TCPKeepAlive yes

When the issue persists despite configuration changes, examine potential network interruptions:

# Check for packet drops during connection
tcpdump -i eth0 'port 22 and (tcp-syn|tcp-ack|tcp-rst)'

For development environments requiring stable long-term connections:

# Using tmux for session persistence
ssh user@host -t 'tmux new -A -s dev_session'

# With automatic reconnection:
while true; do ssh -o ConnectTimeout=5 user@host; sleep 2; done

Some firewall configurations aggressively terminate idle connections. Check for rules affecting SSH timeouts:

# For iptables-based firewalls
sudo iptables -L -n --line-numbers | grep -i timeout

When working with remote servers through SSH, connection stability is crucial for productive development workflows. The "Write failed: Broken pipe" error typically occurs when:

Network inactivity triggers TCP timeout
Intermediate network devices drop idle connections
Server-side SSH daemon configuration has aggressive timeout settings
Client-side keepalive mechanisms aren't properly configured

On your CentOS 5.5 server, edit the SSH daemon configuration:


# sudo vi /etc/ssh/sshd_config

# Add or modify these parameters:

ClientAliveInterval 60

ClientAliveCountMax 3

TCPKeepAlive yes


This configuration tells the server to:

Send keepalive messages every 60 seconds (ClientAliveInterval)
Terminate the connection after 3 failed keepalive attempts (ClientAliveCountMax)
Enable TCP-level keepalive packets (TCPKeepAlive)

For your Ubuntu 11.04 machine, create or modify your SSH client configuration:


# vi ~/.ssh/config

Host *

  ServerAliveInterval 30

  ServerAliveCountMax 2

  TCPKeepAlive yes


This client-side configuration:

Sends keepalive packets every 30 seconds
Gives up after 2 failed attempts
Works alongside server-side settings

For critical development sessions, consider using terminal multiplexers:


# On the server (install if needed)

sudo yum install tmux

# Start a persistent session

tmux new -s dev_session
# Later, to reattach:

tmux attach -t dev_session


When troubleshooting, these commands help identify connection issues:


# Check TCP connection state

netstat -tn | grep ssh

# Monitor network traffic

tcpdump -i eth0 'port 22'
# Test connection quality

mtr -4 -rwc 100 your.server.ip


Ensure your firewall isn't dropping idle connections. For iptables:


# Allow established SSH connections

iptables -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT

# Adjust connection tracking timeout

modprobe ip_conntrack

echo 3600 > /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_established