Blue-Green Upgrade to a New Cluster

When upgrading Elasticsearch itself (for example 7.x → 8.x), it’s safer to use two separate clusters.

Step 1 Create the New (Green) Cluster

Install the new version on new nodes.
Do not mix old and new nodes in the same cluster if the major version is different.

Step 2 Register a Snapshot Repository

On the blue cluster:

PUT _snapshot/backup_repo
{
  "type": "fs",
  "settings": { "location": "/mnt/backup" }
}

Step 3 Take a Snapshot

PUT _snapshot/backup_repo/snapshot_oct2025
{
  "indices": "my_index*",
  "ignore_unavailable": true,
  "include_global_state": false
}

Step 4 Restore to the New Cluster

On the green cluster:

POST _snapshot/backup_repo/snapshot_oct2025/_restore
{
  "indices": "my_index*",
  "rename_pattern": "my_index_v1",
  "rename_replacement": "my_index_v2"
}

Now the green cluster has all data in the upgraded format.

Step 5 Validate and Switch

Point your load balancer or application connection to the new cluster endpoint.
If problems appear, simply point back to the old cluster (blue).

Continuous Sync with Cross-Cluster Replication (CCR)

If your dataset is very large or you cannot afford any delay between blue and green clusters,
Cross-Cluster Replication is the best approach.

CCR keeps a follower index in the new (green) cluster that automatically copies all operations from the leader index in the old (blue) cluster, including new documents, updates, and deletions, in near real-time.

How CCR Works

• Leader Cluster (Blue): keeps the original data
• Follower Cluster (Green): continuously replicates data
• You can later promote the follower cluster to production

CCR uses Elasticsearch’s internal replication engine to send new operations through the remote cluster connection.

Requirements for Cross-Cluster Replication

Requirement	Description
Elasticsearch Platinum or Enterprise license	CCR is a commercial feature (not in basic license)
Both clusters on compatible versions	e.g., 8.5 → 8.10 (same major version)
Remote connection configured	Green must know where to find Blue
Security enabled	TLS/SSL + user roles for replication
Index in Blue must have soft deletes enabled	Default in ES 7+

Check soft deletes setting:

GET my_index_v1/_settings

If missing:

PUT my_index_v1/_settings
{ "index.soft_deletes.enabled": true }

Setting Up CCR

Step 1 Register the Remote Cluster (on Green)

PUT _cluster/settings
{
  "persistent": {
    "cluster.remote.blue_cluster.seeds": [ "blue-node1:9300" ]
  }
}

Step 2 Create a Follower Index

PUT my_index_v2/_ccr/follow
{
  "remote_cluster": "blue_cluster",
  "leader_index": "my_index_v1"
}

Now, my_index_v2 will continuously follow all changes from my_index_v1.

---

Step 3 Monitor Replication

GET my_index_v2/_ccr/stats

Check:

• operations_read → how many operations replicated
• time_since_last_read_millis → lag time
• follower_checkpoint vs leader_checkpoint

---

Step 4 Promote the Follower (Switch to Green)

When ready to make Green the new production cluster:

1. Pause replication: POST my_index_v2/_ccr/unfollow
2. Switch your application endpoint from Blue to Green.
3. Optionally, start replicating back (if needed for rollback).

Now, Green is the active environment, Blue can be kept as a standby or backup.

CCR vs. Reindex/Snapshot

Method	Speed	Real-time	Cross-cluster	License
_reindex	Medium	No	Optional	Free
Snapshot & Restore	Slow	No	Yes	Free
CCR	Fast	Yes	Yes	Paid (Platinum/Enterprise)

CCR is ideal for:

• Real-time migrations
• Large, high-volume clusters
• Zero-downtime cross-region setups