Troubleshooting remote clustersedit

You may encounter several issues when setting up a remote cluster for cross-cluster replication or cross-cluster search.

General troubleshootingedit

Checking whether a remote cluster has connected successfullyedit

A successful call to the cluster settings update API for adding or updating remote clusters does not necessarily mean the configuration is successful. Use the remote cluster info API to verify that a local cluster is successfully connected to a remote cluster.

GET /_remote/info

The API should return "connected" : true. When using API key authentication, it should also return "cluster_credentials": "::es_redacted::".

{
  "cluster_one" : {
    "seeds" : [
      "127.0.0.1:9443"
    ],
    "connected" : true, 
    "num_nodes_connected" : 1,
    "max_connections_per_cluster" : 3,
    "initial_connect_timeout" : "30s",
    "skip_unavailable" : false,
    "cluster_credentials": "::es_redacted::", 
    "mode" : "sniff"
  }
}

The remote cluster has connected successfully.

If present, indicates the remote cluster has connected using API key authentication instead of certificate based authentication.

Enabling the remote cluster serveredit

When using API key authentication, cross-cluster traffic happens on the remote cluster interface, instead of the transport interface. The remote cluster interface is not enabled by default. This means a node is not ready to accept incoming cross-cluster requests by default, while it is ready to send outgoing cross-cluster requests. Ensure you’ve enabled the remote cluster server on every node of the remote cluster. In elasticsearch.yml:

  • Set remote_cluster_server.enabled to true.
  • Configure the bind and publish address for remote cluster server traffic, for example using remote_cluster.host. Without configuring the address, remote cluster traffic may be bound to the local interface, and remote clusters running on other machines can’t connect.
  • Optionally, configure the remote server port using remote_cluster.port (defaults to 9443).

Common issuesedit

The following issues are listed in the order they may occur while setting up a remote cluster.

Remote cluster not reachableedit

Symptomedit

A local cluster may not be able to reach a remote cluster for many reasons. For example, the remote cluster server may not be enabled, an incorrect host or port may be configured, or a firewall may be blocking traffic. When a remote cluster is not reachable, check the logs of the local cluster for a connect_exception.

When the remote cluster is configured using proxy mode:

[2023-06-28T16:36:47,264][WARN ][o.e.t.ProxyConnectionStrategy] [local-node] failed to open any proxy connections to cluster [my]
org.elasticsearch.transport.ConnectTransportException: [][192.168.0.42:9443] connect_exception

When the remote cluster is configured using sniff mode:

[2023-06-28T16:38:37,731][WARN ][o.e.t.SniffConnectionStrategy] [local-node] fetching nodes from external cluster [my] failed
org.elasticsearch.transport.ConnectTransportException: [][192.168.0.42:9443] connect_exception
Resolutionedit
  • Check the host and port for the remote cluster are correct.
  • Ensure the remote cluster server is enabled on the remote cluster.
  • Ensure no firewall is blocking the communication.

TLS trust not establishededit

TLS can be misconfigured on the local or the remote cluster. The result is that the local cluster does not trust the certificate presented by the remote cluster.

Symptomedit

The local cluster logs failed to establish trust with server:

[2023-06-29T09:40:55,465][WARN ][o.e.c.s.DiagnosticTrustManager] [local-node] failed to establish trust with server at [192.168.0.42]; the server provided a certificate with subject name [CN=remote_cluster], fingerprint [529de35e15666ffaa26afa50876a2a48119db03a], no keyUsage and no extendedKeyUsage; the certificate is valid between [2023-01-29T12:08:37Z] and [2032-08-29T12:08:37Z] (current time is [2023-08-16T23:40:55.464275Z], certificate dates are valid); the session uses cipher suite [TLS_AES_256_GCM_SHA384] and protocol [TLSv1.3]; the certificate has subject alternative names [DNS:localhost,DNS:localhost6.localdomain6,IP:127.0.0.1,IP:0:0:0:0:0:0:0:1,DNS:localhost4,DNS:localhost6,DNS:localhost.localdomain,DNS:localhost4.localdomain4,IP:192.168.0.42]; the certificate is issued by [CN=Elastic Auto RemoteCluster CA] but the server did not provide a copy of the issuing certificate in the certificate chain; this ssl context ([(shared) (with trust configuration: JDK-trusted-certs)]) is not configured to trust that issuer but trusts [97] other issuers
sun.security.validator.ValidatorException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target

The remote cluster logs client did not trust this server's certificate:

[2023-06-29T09:40:55,478][WARN ][o.e.x.c.s.t.n.SecurityNetty4Transport] [remote-node] client did not trust this server's certificate, closing connection Netty4TcpChannel{localAddress=/192.168.0.42:9443, remoteAddress=/192.168.0.84:57305, profile=_remote_cluster}
Resolutionedit

Read the warn log message on the local cluster carefully to determine the exact cause of the failure. For example:

  • Is the remote cluster certificate not signed by a trusted CA? This is the most likely cause.
  • Is hostname verification failing?
  • Is the certificate expired?

Once you know the cause, you should be able to fix it by adjusting the remote cluster related SSL settings on either the local cluster or the remote cluster.

Often, the issue is on the local cluster. For example, fix it by configuring necessary trusted CAs (xpack.security.remote_cluster_client.ssl.certificate_authorities).

If you change the elasticsearch.yml file, the associated cluster needs to be restarted for the changes to take effect.

API key authentication issuesedit

Connecting to transport port when using API key authenticationedit

When using API key authentication, a local cluster should connect to a remote cluster’s remote cluster server port (defaults to 9443) instead of the transport port (defaults to 9300). A misconfiguration can lead to a number of symptoms:

Symptom 1edit

It’s recommended to use different CAs and certificates for the transport interface and the remote cluster server interface. If this recommendation is followed, a remote cluster client node does not trust the server certificate presented by a remote cluster on the transport interface.

The local cluster logs failed to establish trust with server:

[2023-06-28T12:48:46,575][WARN ][o.e.c.s.DiagnosticTrustManager] [local-node] failed to establish trust with server at [1192.168.0.42]; the server provided a certificate with subject name [CN=transport], fingerprint [c43e628be2a8aaaa4092b82d78f2bc206c492322], no keyUsage and no extendedKeyUsage; the certificate is valid between [2023-01-29T12:05:53Z] and [2032-08-29T12:05:53Z] (current time is [2023-06-28T02:48:46.574738Z], certificate dates are valid); the session uses cipher suite [TLS_AES_256_GCM_SHA384] and protocol [TLSv1.3]; the certificate has subject alternative names [DNS:localhost,DNS:localhost6.localdomain6,IP:127.0.0.1,IP:0:0:0:0:0:0:0:1,DNS:localhost4,DNS:localhost6,DNS:localhost.localdomain,DNS:localhost4.localdomain4,IP:192.168.0.42]; the certificate is issued by [CN=Elastic Auto Transport CA] but the server did not provide a copy of the issuing certificate in the certificate chain; this ssl context ([xpack.security.remote_cluster_client.ssl (with trust configuration: PEM-trust{/rcs2/ssl/remote-cluster-ca.crt})]) is not configured to trust that issuer, it only trusts the issuer [CN=Elastic Auto RemoteCluster CA] with fingerprint [ba2350661f66e46c746c1629f0c4b645a2587ff4]
sun.security.validator.ValidatorException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target

The remote cluster logs client did not trust this server's certificate:

[2023-06-28T12:48:46,584][WARN ][o.e.x.c.s.t.n.SecurityNetty4Transport] [remote-node] client did not trust this server's certificate, closing connection Netty4TcpChannel{localAddress=/192.168.0.42:9309, remoteAddress=/192.168.0.84:60810, profile=default}
Symptom 2edit

The CA and certificate can be shared between the transport and remote cluster server interface. Since a remote cluster client does not have a client certificate by default, the server will fail to verify the client certificate.

The local cluster logs Received fatal alert: bad_certificate:

[2023-06-28T12:43:30,705][WARN ][o.e.t.TcpTransport       ] [local-node] exception caught on transport layer [Netty4TcpChannel{localAddress=/192.168.0.84:60738, remoteAddress=/192.168.0.42:9309, profile=_remote_cluster}], closing connection
io.netty.handler.codec.DecoderException: javax.net.ssl.SSLHandshakeException: Received fatal alert: bad_certificate

The remote cluster logs Empty client certificate chain:

[2023-06-28T12:43:30,772][WARN ][o.e.t.TcpTransport       ] [remote-node] exception caught on transport layer [Netty4TcpChannel{localAddress=/192.168.0.42:9309, remoteAddress=/192.168.0.84:60783, profile=default}], closing connection
io.netty.handler.codec.DecoderException: javax.net.ssl.SSLHandshakeException: Empty client certificate chain
Symptom 3edit

If the remote cluster client is configured for mTLS and provides a valid client certificate, the connection fails because the client does not send the expected authentication header.

The local cluster logs missing authentication:

[2023-06-28T13:04:52,710][WARN ][o.e.t.ProxyConnectionStrategy] [local-node] failed to open any proxy connections to cluster [my]
org.elasticsearch.transport.RemoteTransportException: [remote-node][192.168.0.42:9309][cluster:internal/remote_cluster/handshake]
Caused by: org.elasticsearch.ElasticsearchSecurityException: missing authentication credentials for action [cluster:internal/remote_cluster/handshake]

This does not show up in the logs of the remote cluster.

Symptom 4edit

If anonymous access is enabled on the remote cluster and it does not require authentication, depending on the privileges of the anonymous user, the local cluster may log the following.

If the anonymous user does not the have necessary privileges to make a connection, the local cluster logs unauthorized:

org.elasticsearch.transport.RemoteTransportException: [remote-node][192.168.0.42:9309][cluster:internal/remote_cluster/handshake]
Caused by: org.elasticsearch.ElasticsearchSecurityException: action [cluster:internal/remote_cluster/handshake] is unauthorized for user [anonymous_foo] with effective roles [reporting_user], this action is granted by the cluster privileges [cross_cluster_search,cross_cluster_replication,manage,all]

If the anonymous user has necessary privileges, for example it is a superuser, the local cluster logs requires channel profile to be [_remote_cluster], but got [default]:

[2023-06-28T13:09:52,031][WARN ][o.e.t.ProxyConnectionStrategy] [local-node] failed to open any proxy connections to cluster [my]
org.elasticsearch.transport.RemoteTransportException: [remote-node][192.168.0.42:9309][cluster:internal/remote_cluster/handshake]
Caused by: java.lang.IllegalArgumentException: remote cluster handshake action requires channel profile to be [_remote_cluster], but got [default]
Resolutionedit

Check the port number and ensure you are indeed connecting to the remote cluster server instead of the transport interface.

Connecting without a cross-cluster API keyedit

A local cluster uses the presence of a cross-cluster API key to determine the model with which it connects to a remote cluster. If a cross-cluster API key is present, it uses API key based authentication. Otherwise, it uses certificate based authentication. You can check what model is being used with the remote cluster info API on the local cluster:

GET /_remote/info

The API should return "connected" : true. When using API key authentication, it should also return "cluster_credentials": "::es_redacted::".

{
  "cluster_one" : {
    "seeds" : [
      "127.0.0.1:9443"
    ],
    "connected" : true, 
    "num_nodes_connected" : 1,
    "max_connections_per_cluster" : 3,
    "initial_connect_timeout" : "30s",
    "skip_unavailable" : false,
    "cluster_credentials": "::es_redacted::", 
    "mode" : "sniff"
  }
}

The remote cluster has connected successfully.

If present, indicates the remote cluster has connected using API key authentication instead of certificate based authentication.

Besides checking the response of the remote cluster info API, you can also check the logs.

Symptom 1edit

If no cross-cluster API key is used, the local cluster uses the certificate based authentication method, and connects to the remote cluster using the TLS configuration of the transport interface. If the remote cluster has different TLS CA and certificate for transport and remote cluster server interfaces (which is the recommendation), TLS verification will fail.

The local cluster logs failed to establish trust with server:

[2023-06-28T12:51:06,452][WARN ][o.e.c.s.DiagnosticTrustManager] [local-node] failed to establish trust with server at [<unknown host>]; the server provided a certificate with subject name [CN=remote_cluster], fingerprint [529de35e15666ffaa26afa50876a2a48119db03a], no keyUsage and no extendedKeyUsage; the certificate is valid between [2023-01-29T12:08:37Z] and [2032-08-29T12:08:37Z] (current time is [2023-06-28T02:51:06.451581Z], certificate dates are valid); the session uses cipher suite [TLS_AES_256_GCM_SHA384] and protocol [TLSv1.3]; the certificate has subject alternative names [DNS:localhost,DNS:localhost6.localdomain6,IP:127.0.0.1,IP:0:0:0:0:0:0:0:1,DNS:localhost4,DNS:localhost6,DNS:localhost.localdomain,DNS:localhost4.localdomain4,IP:192.168.0.42]; the certificate is issued by [CN=Elastic Auto RemoteCluster CA] but the server did not provide a copy of the issuing certificate in the certificate chain; this ssl context ([xpack.security.transport.ssl (with trust configuration: PEM-trust{/rcs2/ssl/transport-ca.crt})]) is not configured to trust that issuer, it only trusts the issuer [CN=Elastic Auto Transport CA] with fingerprint [bbe49e3f986506008a70ab651b188c70df104812]
sun.security.validator.ValidatorException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target

The remote cluster logs client did not trust this server's certificate:

[2023-06-28T12:52:16,914][WARN ][o.e.x.c.s.t.n.SecurityNetty4Transport] [remote-node] client did not trust this server's certificate, closing connection Netty4TcpChannel{localAddress=/192.168.0.42:9443, remoteAddress=/192.168.0.84:60981, profile=_remote_cluster}
Symptom 2edit

Even if TLS verification is not an issue, the connection fails due to missing credentials.

The local cluster logs Please ensure you have configured remote cluster credentials:

Caused by: java.lang.IllegalArgumentException: Cross cluster requests through the dedicated remote cluster server port require transport header [_cross_cluster_access_credentials] but none found. Please ensure you have configured remote cluster credentials on the cluster originating the request.

This does not show up in the logs of the remote cluster.

Resolutionedit

Add the cross-cluster API key to Elasticsearch keystore on every node of the local cluster. Restart the local cluster to reload the keystore.

Using the wrong API key typeedit

API key based authentication requires cross-cluster API keys. It does not work with REST API keys.

Symptomedit

The local cluster logs authentication expected API key type of [cross_cluster]:

[2023-06-28T13:26:53,962][WARN ][o.e.t.ProxyConnectionStrategy] [local-node] failed to open any proxy connections to cluster [my]
org.elasticsearch.transport.RemoteTransportException: [remote-node][192.168.0.42:9443][cluster:internal/remote_cluster/handshake]
Caused by: org.elasticsearch.ElasticsearchSecurityException: authentication expected API key type of [cross_cluster], but API key [agZXJocBmA2beJfq2yKu] has type [rest]

This does not show up in the logs of the remote cluster.

Resolutionedit

Ask the remote cluster administrator to create and distribute a cross-cluster API key. Replace the existing API key in the Elasticsearch keystore with this cross-cluster API key on every node of the local cluster. Restart the local cluster for keystore changes to take effect.

Invalid API keyedit

A cross-cluster API can fail to authenticate. For example, when its credentials are incorrect, or if it’s invalidated or expired.

Symptomedit

The local cluster logs unable to authenticate:

[2023-06-28T13:22:58,264][WARN ][o.e.t.ProxyConnectionStrategy] [local-node] failed to open any proxy connections to cluster [my]
org.elasticsearch.transport.RemoteTransportException: [remote-node][192.168.0.42:9443][cluster:internal/remote_cluster/handshake]
Caused by: org.elasticsearch.ElasticsearchSecurityException: unable to authenticate user [agZXJocBmA2beJfq2yKu] for action [cluster:internal/remote_cluster/handshake]

The remote cluster logs Authentication using apikey failed:

[2023-06-28T13:24:38,744][WARN ][o.e.x.s.a.ApiKeyAuthenticator] [remote-node] Authentication using apikey failed - invalid credentials for API key [agZXJocBmA2beJfq2yKu]
Resolutionedit

Ask the remote cluster administrator to create and distribute a cross-cluster API key. Replace the existing API key in the Elasticsearch keystore with this cross-cluster API key on every node of the local cluster. Restart the local cluster for keystore changes to take effect.

API key or local user has insufficient privilegesedit

The effective permission for a local user running requests on a remote cluster is determined by the intersection of the cross-cluster API key’s privileges and the local user’s remote_indices privileges.

Symptomedit

Request failures due to insufficient privileges result in API responses like:

{
    "type": "security_exception",
    "reason": "action [indices:data/read/search] towards remote cluster is unauthorized for user [foo] with assigned roles [foo-role] authenticated by API key id [agZXJocBmA2beJfq2yKu] of user [elastic-admin] on indices [cd], this action is granted by the index privileges [read,all]"
}

This does not show up in any logs.

Resolutionedit
  1. Check that the local user has the necessary remote_indices privileges. Grant sufficient remote_indices privileges if necessary.
  2. If permission is not an issue locally, ask the remote cluster administrator to create and distribute a cross-cluster API key. Replace the existing API key in the Elasticsearch keystore with this cross-cluster API key on every node of the local cluster. Restart the local cluster for keystore changes to take effect.

Local user has no remote_indices privilegesedit

This is a special case of insufficient privileges. In this case, the local user has no remote_indices privileges at all for the target remote cluster. Elasticsearch can detect that and issue a more explicit error response.

Symptomedit

This results in API responses like:

{
    "type": "security_exception",
    "reason": "action [indices:data/read/search] towards remote cluster [my] is unauthorized for user [foo] with effective roles [] (assigned roles [foo-role] were not found) because no remote indices privileges apply for the target cluster"
}
Resolutionedit

Grant sufficient remote_indices privileges to the local user.