Hadoop启动报Connection refused错误，需重复格式化NameNode求助

I’ve run into this exact frustrating issue before—having to reformat the NameNode every time you start Hadoop gets old fast. Let’s break down the most common causes and step-by-step fixes:

1. Check Configuration Files For Persistent Storage Paths #

The #1 reason this happens is that your NameNode is configured to store metadata in a temporary directory (like /tmp) which gets wiped when your system reboots. Here’s how to fix it:

• Open your hdfs-site.xml file (usually in $HADOOP_HOME/etc/hadoop).
• Look for the dfs.namenode.name.dir property. If it points to a /tmp path, replace it with a persistent directory, e.g.:

  <property>
      <name>dfs.namenode.name.dir</name>
      <value>/opt/hadoop/data/namenode</value>
  </property>
  

• Verify core-site.xml has the correct NameNode address set:

  <property>
      <name>fs.defaultFS</name>
      <value>hdfs://localhost:9000</value>
  </property>
  

2. Create The Persistent Directory & Set Permissions #

Once you’ve updated the config, create the directory and ensure Hadoop has access to it:

# Create the directory (match the path from hdfs-site.xml)
mkdir -p /opt/hadoop/data/namenode

# Change ownership to your Hadoop user (replace "hadoop" with your actual username)
sudo chown -R hadoop:hadoop /opt/hadoop/data

# Set proper read/write permissions
chmod -R 755 /opt/hadoop/data

3. Format The NameNode One Final Time #

After updating configs and setting up the directory, format the NameNode once—you shouldn’t need to do this again unless you intentionally reset HDFS:

hadoop namenode -format

4. Verify Services Start Correctly #

Start Hadoop properly and check if all required daemons are running:

# Start HDFS and YARN services
start-dfs.sh
start-yarn.sh

# List running Hadoop processes
jps

You should see NameNode, DataNode, SecondaryNameNode, ResourceManager, and NodeManager listed. If any are missing, check their log files in $HADOOP_HOME/logs for specific errors.

5. Troubleshoot With Logs If Issues Persist #

If you still get connection errors, dive into the NameNode log (e.g., hadoop-*-namenode-<your-hostname>.log in the logs folder). Search for keywords like Connection refused or Failed to load metadata—this will pinpoint exactly what’s going wrong.

内容的提问来源于stack exchange，提问作者Amol