Hey Ive,

Replacing a dead node is a 3-part process.Note that the following procedure is for the standard 2 node failover setup. Custom clusters might need a different process depending on their configuration.

• Part 1: prepare the new node
  • Install cassandra
    1. Place the cassandra files under C:\Program files\cassandra (Currently the best way is to copy them from the other server, DO NO copy the "c:\program files\data" folder)
    2. Go to c:\program files\cassandra\conf\cassandra.yaml
    3. Set the following parameters
    • cluster_name: (name of the dataminer cluster !must be the same as the other node!)
    • data_file_directories: location to save data
    • seeds:Ip of this node and other node
    • disk_optimization_strategy: ssd or spinning
    • listen_address: Ip of this node
    • broadcast_rpc_address: IP of this node
    1. Save and close the file
    2. Open an elevated powershell prompt
    3. Navigate to C:\program files\cassandra\bin
    4. Execute $env:CASSANDRA_HOME = ‘C:\progra~1\Cassandra\’
    5. Execute $env:JAVA_HOME = ‘C:\progra~1\Cassandra\Java\’
    6. Execute cassandra.ps1 –install
    7. Open port 9042 and 7000 on the firewall
  • In the registry under "Computer\HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Apache Software Foundation\Procrun 2.0\cassandra\Parameters\Java" add "-Dcassandra.replace_address_first_boot=<dead_node_ip>" to the Options-key
  • In cassandra.yaml add or set "auto_bootstrap:true"
• Part 2: Updating the old node
  • In the cassandra.yaml under "-seeds:" replace the ip of the dead node with the ip of the new one.
  • Restart the node
• Part 3: Start the new node
  • The new node should automatically start bootstrapping the data: this can be followed with nodetool netstats
  • If the node was down longer than the max_hint_window_in_ms. then you must also run a nodetool repair

Hope this helps, if there any questions feel free to leave a comment.