Getting started with cross-cluster replication

With cross-cluster replication, you index data to a leader index, and OpenSearch replicates that data to one or more read-only follower indexes. All subsequent operations on the leader are replicated on the follower, such as creating, updating, or deleting documents.

Prerequisites

Cross-cluster replication has the following prerequisites:

  • Both the leader and follower cluster must have the replication plugin installed.
  • If you’ve overridden node.roles in opensearch.yml on the follower cluster, make sure it also includes the remote_cluster_client role:

    1. node.roles: [<other_roles>, remote_cluster_client]

Permissions

Make sure the Security plugin is either enabled on both clusters or disabled on both clusters. If you disabled the Security plugin, you can skip this section. However, we strongly recommend enabling the Security plugin in production scenarios.

If the Security plugin is enabled, make sure that non-admin users are mapped to the appropriate permissions so they can perform replication actions. For index and cluster-level permissions requirements, see Cross-cluster replication permissions.

In addition, verify and add the distinguished names (DNs) of each follower cluster node on the leader cluster to allow connections from the followers to the leader.

First, get the node’s DN from each follower cluster:

  1. curl -XGET -k -u 'admin:admin' 'https://localhost:9200/_opendistro/_security/api/ssl/certs?pretty'
  2. {
  3. "transport_certificates_list": [
  4. {
  5. "issuer_dn" : "CN=Test,OU=Server CA 1B,O=Test,C=US",
  6. "subject_dn" : "CN=follower.test.com", # To be added under leader's nodes_dn configuration
  7. "not_before" : "2021-11-12T00:00:00Z",
  8. "not_after" : "2022-12-11T23:59:59Z"
  9. }
  10. ]
  11. }

Then verify that it’s part of the leader cluster configuration in opensearch.yml. Otherwise, add it under the following setting:

  1. plugins.security.nodes_dn:
  2. - "CN=*.leader.com, OU=SSL, O=Test, L=Test, C=DE" # Already part of the configuration
  3. - "CN=follower.test.com" # From the above response from follower

Example setup

To start two single-node clusters on the same network, save this sample file as docker-compose.yml and run docker-compose up:

  1. version: '3'
  2. services:
  3. replication-node1:
  4. image: opensearchproject/opensearch:2.7.0
  5. container_name: replication-node1
  6. environment:
  7. - cluster.name=leader-cluster
  8. - discovery.type=single-node
  9. - bootstrap.memory_lock=true
  10. - "OPENSEARCH_JAVA_OPTS=-Xms512m -Xmx512m"
  11. ulimits:
  12. memlock:
  13. soft: -1
  14. hard: -1
  15. volumes:
  16. - opensearch-data2:/usr/share/opensearch/data
  17. ports:
  18. - 9201:9200
  19. - 9700:9600 # required for Performance Analyzer
  20. networks:
  21. - opensearch-net
  22. replication-node2:
  23. image: opensearchproject/opensearch:2.7.0
  24. container_name: replication-node2
  25. environment:
  26. - cluster.name=follower-cluster
  27. - discovery.type=single-node
  28. - bootstrap.memory_lock=true
  29. - "OPENSEARCH_JAVA_OPTS=-Xms512m -Xmx512m"
  30. ulimits:
  31. memlock:
  32. soft: -1
  33. hard: -1
  34. volumes:
  35. - opensearch-data1:/usr/share/opensearch/data
  36. ports:
  37. - 9200:9200
  38. - 9600:9600 # required for Performance Analyzer
  39. networks:
  40. - opensearch-net
  41. volumes:
  42. opensearch-data1:
  43. opensearch-data2:
  44. networks:
  45. opensearch-net:

After the clusters start, verify the names of each:

  1. curl -XGET -u 'admin:admin' -k 'https://localhost:9201'
  2. {
  3. "cluster_name" : "leader-cluster",
  4. ...
  5. }
  6. curl -XGET -u 'admin:admin' -k 'https://localhost:9200'
  7. {
  8. "cluster_name" : "follower-cluster",
  9. ...
  10. }

For this example, use port 9201 (replication-node1) as the leader and port 9200 (replication-node2) as the follower cluster.

To get the IP address for the leader cluster, first identify its container ID:

  1. docker ps
  2. CONTAINER ID IMAGE PORTS NAMES
  3. 3b8cdc698be5 opensearchproject/opensearch:2.7.0 0.0.0.0:9200->9200/tcp, 0.0.0.0:9600->9600/tcp, 9300/tcp replication-node2
  4. 731f5e8b0f4b opensearchproject/opensearch:2.7.0 9300/tcp, 0.0.0.0:9201->9200/tcp, 0.0.0.0:9700->9600/tcp replication-node1

Then get that container’s IP address:

  1. docker inspect --format='{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}' 731f5e8b0f4b
  2. 172.22.0.3

Set up a cross-cluster connection

Cross-cluster replication follows a “pull” model, so most changes occur on the follower cluster, not the leader cluster.

On the follower cluster, add the IP address (with port 9300) for each seed node. Because this is a single-node cluster, you only have one seed node. Provide a descriptive name for the connection, which you’ll use in the request to start replication:

  1. curl -XPUT -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9200/_cluster/settings?pretty' -d '
  2. {
  3. "persistent": {
  4. "cluster": {
  5. "remote": {
  6. "my-connection-alias": {
  7. "seeds": ["172.22.0.3:9300"]
  8. }
  9. }
  10. }
  11. }
  12. }'

Start replication

To get started, create an index called leader-01 on the leader cluster:

  1. curl -XPUT -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9201/leader-01?pretty'

Then start replication from the follower cluster. In the request body, provide the connection name and leader index that you want to replicate, along with the security roles you want to use:

  1. curl -XPUT -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_start?pretty' -d '
  2. {
  3. "leader_alias": "my-connection-alias",
  4. "leader_index": "leader-01",
  5. "use_roles":{
  6. "leader_cluster_role": "all_access",
  7. "follower_cluster_role": "all_access"
  8. }
  9. }'

If the Security plugin is disabled, omit the use_roles parameter. If it’s enabled, however, you must specify the leader and follower cluster roles that OpenSearch will use to authenticate the request. This example uses all_access for simplicity, but we recommend creating a replication user on each cluster and mapping it accordingly.

This command creates an identical read-only index named follower-01 on the follower cluster that continuously stays updated with changes to the leader-01 index on the leader cluster. Starting replication creates a follower index from scratch – you can’t convert an existing index to a follower index.

Confirm replication

After replication starts, get the status:

  1. curl -XGET -k -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_status?pretty'
  2. {
  3. "status" : "SYNCING",
  4. "reason" : "User initiated",
  5. "leader_alias" : "my-connection-alias",
  6. "leader_index" : "leader-01",
  7. "follower_index" : "follower-01",
  8. "syncing_details" : {
  9. "leader_checkpoint" : -1,
  10. "follower_checkpoint" : -1,
  11. "seq_no" : 0
  12. }
  13. }

Possible statuses are SYNCING, BOOTSTRAPPING, PAUSED, and REPLICATION NOT IN PROGRESS.

The leader and follower checkpoint values begin as negative numbers and reflect the shard count (-1 for one shard, -5 for five shards, and so on). The values increment with each change and illustrate how many updates the follower is behind the leader. If the indexes are fully synced, the values are the same.

To confirm that replication is actually happening, add a document to the leader index:

  1. curl -XPUT -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9201/leader-01/_doc/1?pretty' -d '{"The Shining": "Stephen King"}'

Then validate the replicated content on the follower index:

  1. curl -XGET -k -u 'admin:admin' 'https://localhost:9200/follower-01/_search?pretty'
  2. {
  3. ...
  4. "hits": [{
  5. "_index": "follower-01",
  6. "_id": "1",
  7. "_score": 1.0,
  8. "_source": {
  9. "The Shining": "Stephen King"
  10. }
  11. }]
  12. }

Pause and resume replication

You can temporarily pause replication of an index if you need to remediate issues or reduce load on the leader cluster:

  1. curl -XPOST -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_pause?pretty' -d '{}'

To confirm that replication is paused, get the status:

  1. curl -XGET -k -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_status?pretty'
  2. {
  3. "status" : "PAUSED",
  4. "reason" : "User initiated",
  5. "leader_alias" : "my-connection-alias",
  6. "leader_index" : "leader-01",
  7. "follower_index" : "follower-01"
  8. }

When you’re done making changes, resume replication:

  1. curl -XPOST -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_resume?pretty' -d '{}'

When replication resumes, the follower index picks up any changes that were made to the leader index while replication was paused.

Note that you can’t resume replication after it’s been paused for more than 12 hours. You must stop replication, delete the follower index, and restart replication of the leader.

Stop replication

When you no longer need to replicate an index, terminate replication from the follower cluster:

  1. curl -XPOST -k -H 'Content-Type: application/json' -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_stop?pretty' -d '{}'

When you stop replication, the follower index un-follows the leader and becomes a standard index that you can write to. You can’t restart replication after stopping it.

Get the status to confirm that the index is no longer being replicated:

  1. curl -XGET -k -u 'admin:admin' 'https://localhost:9200/_plugins/_replication/follower-01/_status?pretty'
  2. {
  3. "status" : "REPLICATION NOT IN PROGRESS"
  4. }

You can further confirm that replication is stopped by making modifications to the leader index and confirming they don’t show up on the follower index.