Replacing a dead seed node
Steps to replace a seed node.
Because Cassandra doesn't allow a seed node to be bootstrapped, additional steps must be performed when replacing a dead node. The methods presented in this topic augment the steps described in Replacing a dead node.
Choose the method that best suits your needs:
Method 1: Replace the node and then run nodetool repair.
When you start the replacement node, Cassandra displays a warning stating that the node did not auto bootstrap. The warning is expected because Cassandra will ignore the auto_bootstrap properties of the replacement node when it's listed the seed node list.
Method 2: Use a different IP address for the replacement node.
Note: You don't have to remove the old seed address, no problems occur when a seed address exists for a node that isn't running.
- When starting the node, do not use the replace_address option.
- Add the replacement node's IP address to the cluster's seed list.
Method 3: Before adding the replacement seed node:
- Remove the existing seed node's IP from the cluster's seed list.
- Perform a rolling restart of all the other nodes in the cluster to ensure the complete removal of the seed node from the gossip process.
- Add the new seed node and let it auto bootstrap.
Add the replacement seed node's IP to all the node's seed lists.
You do not need to restart the nodes.