How to set Apache Cassandra seed nodes

Apache Cassandra nodes need reachable seed addresses when they start so gossip can discover the rest of the ring. If the list points at retired hosts, unreachable networks, or a joining non-seed node points only at itself, startup can stall and cluster checks may show missing or down peers.

The seed list lives under seed_provider in /etc/cassandra/cassandra.yaml on packaged Linux installs. SimpleSeedProvider expects one comma-delimited string of addresses, and Cassandra uses those addresses as gossip contact points rather than as replica owners or client connection targets.

Use existing UN nodes as seeds, normally more than one per datacenter or failure domain when available. Change the file on each Cassandra node and restart nodes one at a time so at least one seed remains reachable while the cluster gossips the updated view.

Steps to set Apache Cassandra seed nodes:

Check the current ring from a running Cassandra node.

$ nodetool status
Datacenter: DC1
==============
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address       Load       Tokens  Owns  Host ID                               Rack
UN  10.10.20.11   842.31 MiB 16      ?     8f87a6f2-3bb0-4d09-b63c-5d6f6b9c45d1  RAC1
UN  10.10.20.12   819.44 MiB 16      ?     67d2b72a-4d55-42c7-a85f-c662b08ec8e8  RAC1
UN  10.10.20.13   807.02 MiB 16      ?     b7b90a0a-391a-420d-97c6-a34a2cf97f9f  RAC1

Choose seed candidates that already show UN. Fix down nodes or topology mismatches before using them as seed contact points.

Choose the seed addresses for the cluster.

Use stable private addresses or resolvable hostnames that Cassandra nodes can reach on the internode storage port. In a multi-datacenter cluster, keep at least one reachable seed in each datacenter; using two per datacenter is common when capacity allows it.

Back up the Cassandra YAML file on the first node to change.

$ sudo cp -a /etc/cassandra/cassandra.yaml /etc/cassandra/cassandra.yaml.bak

Open the Cassandra YAML file.

$ sudoedit /etc/cassandra/cassandra.yaml

Set the SimpleSeedProvider seed list.
```
seed_provider:
  - class_name: org.apache.cassandra.locator.SimpleSeedProvider
    parameters:
      - seeds: "10.10.20.11,10.10.20.12"
```
The same seed list can be used on every node in the datacenter. When a new server should become a seed, bootstrap it first with existing seeds, wait until it is UN, then add it to the seed list during a later rolling restart.

Verify the saved seed value.

$ sudo grep -n -- "- seeds:" /etc/cassandra/cassandra.yaml
706:      - seeds: "10.10.20.11,10.10.20.12"

Restart Cassandra on only the edited node.
```
$ sudo systemctl restart cassandra
```
Do not restart every node at once. Wait for the edited node to rejoin before changing the next node so seed contact points remain available.
Confirm gossip is running after the restart.
```
$ nodetool statusgossip
running
```

Confirm the restarted node is up and normal in the cluster view.

$ nodetool status
Datacenter: DC1
==============
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address       Load       Tokens  Owns  Host ID                               Rack
UN  10.10.20.11   842.93 MiB 16      ?     8f87a6f2-3bb0-4d09-b63c-5d6f6b9c45d1  RAC1
UN  10.10.20.12   819.44 MiB 16      ?     67d2b72a-4d55-42c7-a85f-c662b08ec8e8  RAC1
UN  10.10.20.13   807.02 MiB 16      ?     b7b90a0a-391a-420d-97c6-a34a2cf97f9f  RAC1

The edited node should return as UN. If it stays down or missing, check Cassandra logs for seed reachability, listen address, and internode port errors.

Repeat the edit and restart on each remaining Cassandra node.

Keep the seed list consistent across nodes unless a node has a documented network reason to use a different contact point.