How to fix ShardNotFoundException: no shard available for search in Elasticsearch

Use the cluster health API to see the overall status:

curl -X GET "localhost:9200/_cluster/health?pretty"

Look at the response:
- status: red (primary shards missing), yellow (replicas missing), or green (all OK)
- unassigned_shards: count of shards not allocated to any node
- active_shards_percent_as_number: percentage of active shards

Then view detailed shard status:

curl -X GET "localhost:9200/_cat/shards?v"

Find shards with state UNASSIGNED and note the index and shard number.

This API reveals exactly why a shard is unassigned:

curl -X GET "localhost:9200/_cluster/allocation/explain?pretty" \
  -H 'Content-Type: application/json' \
  -d '{
    "index": "your_index_name",
    "shard": 0,
    "primary": true
  }'

The response includes:
- can_allocate: whether the shard can be assigned
- allocate_explanation: human-readable reason (e.g., "no nodes available after filtering")
- unassigned_info: reason shard became unassigned (e.g., NODE_LEFT, CLUSTER_RECOVERED)

Common reasons:
- "node does not meet allocation deciders" — incompatible disk/memory
- "no nodes available after filtering" — insufficient replicas for configured count
- "cannot allocate to same node as another copy" — all eligible nodes are taken

Check if shard allocation has been disabled (common during rolling restarts):

curl -X GET "localhost:9200/_cluster/settings?pretty"

Look for cluster.routing.allocation.enable setting. If it is set to "none" or "primaries", enable it:

curl -X PUT "localhost:9200/_cluster/settings" \
  -H 'Content-Type: application/json' \
  -d '{
    "persistent": {
      "cluster.routing.allocation.enable": "all"
    }
  }'

Wait a few seconds and check cluster health again. Shards should begin allocating.

If you have more replicas configured than available nodes, replicas cannot allocate.

Check current settings:

curl -X GET "localhost:9200/your_index_name/_settings?pretty"

Look for number_of_replicas. If you have fewer nodes than replicas + 1 (primary), reduce replicas:

curl -X PUT "localhost:9200/your_index_name/_settings" \
  -H 'Content-Type: application/json' \
  -d '{
    "index": {
      "number_of_replicas": 0
    }
  }'

For production, the better long-term fix is to add more nodes to your cluster so replicas can spread across them for redundancy.

Elasticsearch will not allocate shards to nodes running low on disk or memory:

curl -X GET "localhost:9200/_nodes/stats/fs,jvm?pretty"

Look at each node's:
- fs.available_in_bytes: free disk space
- jvm.mem.heap_used_percent: heap memory pressure

If disk is near capacity, either:
1. Delete old indices: curl -X DELETE "localhost:9200/old_index"
2. Add more disk space to the node
3. Increase the low disk watermark thresholds (advanced; not recommended long-term)

If heap is full, increase Xmx in elasticsearch.yml and restart the node.

If you have data loss and need to force allocation of unassigned primary shards (risky—may cause inconsistency):

curl -X POST "localhost:9200/_cluster/reroute" \
  -H 'Content-Type: application/json' \
  -d '{
    "commands": [{
      "allocate_stale_primary": {
        "index": "your_index_name",
        "shard": 0,
        "node": "node_id"
      }
    }]
  }'

Only do this if: Primary shard data is lost, all replicas are gone, and you accept potential data inconsistency. This should never be done in production without careful analysis.

Check node IDs with:

curl -X GET "localhost:9200/_cat/nodes?v"

After making changes, monitor the cluster as shards reallocate:

curl -X GET "localhost:9200/_cluster/health?wait_for_status=green&timeout=5m&pretty"

This waits up to 5 minutes for the cluster to reach green status. Watch the logs:

tail -f /var/log/elasticsearch/elasticsearch.log | grep -i "allocation"

You'll see messages like "completed allocation of shard [index_name][0]". Once status is green and unassigned_shards is 0, your searches should work again.