Elasticsearch 503错误求助：MasterNotDiscoveredException问题未解决

Hey there, let's work through this Elasticsearch MasterNotDiscoveredException (the 503 error you're seeing) together. I’ve run into this issue plenty of times, so here are some actionable checks and fixes to try out:

• Validate core cluster configuration in elasticsearch.yml
  This is the most common culprit. Double-check these key settings:

  • Ensure node.master: true is set on at least one node (mandatory for single-node clusters). If you’re running multiple nodes, confirm at least three master-eligible nodes for high availability.
  • For multi-node clusters, discovery.seed_hosts should list all master-eligible nodes’ IPs or hostnames (e.g., discovery.seed_hosts: ["10.0.0.2", "10.0.0.3", "10.0.0.4"]). For single-node setups, set this to ["localhost"] or remove the entry entirely.
  • The cluster.initial_master_nodes setting should only be configured once when initializing the cluster. It must include the names of all master-eligible nodes (e.g., cluster.initial_master_nodes: ["node-1", "node-2", "node-3"]). For a single node, use your node’s name (set via node.name: my-single-node if needed).

• Check network connectivity (multi-node clusters)

  • Confirm all nodes can communicate over port 9300 (used for inter-node cluster communication). Firewalls, security groups, or VPNs often block this port by mistake.
  • Test connectivity with tools like nc: run nc -zv <node-ip> 9300 from each node to verify the port is open.

• Inspect Elasticsearch logs for detailed errors
  Head to your Elasticsearch logs directory (typically <install-dir>/logs/elasticsearch.log) and look for clues. Common red flags include:

  • Mismatched cluster.name across nodes (all nodes must share the same cluster name)
  • Permission issues on the data or logs directories (Elasticsearch needs read/write access)
  • Node name conflicts or invalid hostnames in discovery settings

• Single-node cluster quick fixes
  If you’re running a single node, avoid multi-node-specific settings that can confuse the cluster:

  • Remove any unused discovery.seed_hosts entries that point to non-existent nodes
  • Ensure both node.master and node.data are set to true (this is the default, but it’s worth verifying)

• Last resort: Reset cluster state (test clusters only)
  If you’re working with a test cluster and don’t have critical data, you can reset the cluster state:

  1. Stop all Elasticsearch nodes
  2. Delete the data directory (usually <install-dir>/data/ – this will erase all your data)
  3. Restart the node(s) with the corrected elasticsearch.yml settings

After trying these steps, run curl http://localhost:9200/_cluster/health?pretty to check the cluster health. You should see a status of green or yellow once the master node is discovered.

内容的提问来源于stack exchange，提问作者Anand