Different cluster node names

[mxmcarl]

Hi,
I built a cluster of three nodes in Erlang/OTP 24 and Rabbitmq V3.8.17.But they have different cluster names.The ways to build clusters are peer discovery(rabbitmq_peer_discovery_common Plugin)

[root@host1 ~]# rabbitmqctl eval "rabbit_mnesia:cluster_nodes(all)."
['[email protected]','[email protected]',
'[email protected]']
[root@host1 ~]# rabbitmqctl eval "rabbit_nodes:cluster_name()."
<<"[email protected]">>
[root@host1 ~]#

[root@host2 ~]# rabbitmqctl eval "rabbit_mnesia:cluster_nodes(all)."
['[email protected]','[email protected]',
'[email protected]']
[root@host2 ~]# rabbitmqctl eval "rabbit_nodes:cluster_name()."
<<"[email protected]">>
[root@host2 ~]#

[root@host3 ~]# rabbitmqctl eval "rabbit_mnesia:cluster_nodes(all)."
['[email protected]','[email protected]',
'[email protected]']
[root@host3 ~]# rabbitmqctl eval "rabbit_nodes:cluster_name()."
<<"[email protected]">>
[root@host3 ~]#

[root@host3 ~]# rabbitmqctl cluster_status
Cluster status of node [email protected] ...
Basics

Cluster name: [email protected]

Disk Nodes

[email protected]
[email protected]
[email protected]

Running Nodes

[email protected]
[email protected]
[email protected]

Versions

[email protected]: RabbitMQ 3.8.17 on Erlang 24.0.2
[email protected]: RabbitMQ 3.8.17 on Erlang 24.0.2
[email protected]: RabbitMQ 3.8.17 on Erlang 24.0.2

Maintenance status

Node: [email protected], status: not under maintenance
Node: [email protected], status: not under maintenance
Node: [email protected], status: not under maintenance

Alarms

(none)

Network Partitions

(none)

Listeners

Node: [email protected], interface: [::], port: 15672, protocol: http, purpose: HTTP API
Node: [email protected], interface: [::], port: 61613, protocol: stomp, purpose: STOMP
Node: [email protected], interface: [::], port: 15674, protocol: http/web-stomp, purpose: STOMP over WebSockets
Node: [email protected], interface: [::], port: 1883, protocol: mqtt, purpose: MQTT
Node: [email protected], interface: [::], port: 25672, protocol: clustering, purpose: inter-node and CLI tool communication
Node: [email protected], interface: [::], port: 5672, protocol: amqp, purpose: AMQP 0-9-1 and AMQP 1.0
Node: [email protected], interface: [::], port: 15674, protocol: http/web-stomp, purpose: STOMP over WebSockets
Node: [email protected], interface: [::], port: 25672, protocol: clustering, purpose: inter-node and CLI tool communication
Node: [email protected], interface: [::], port: 15672, protocol: http, purpose: HTTP API
Node: [email protected], interface: [::], port: 61613, protocol: stomp, purpose: STOMP
Node: [email protected], interface: [::], port: 1883, protocol: mqtt, purpose: MQTT
Node: [email protected], interface: [::], port: 5672, protocol: amqp, purpose: AMQP 0-9-1 and AMQP 1.0
Node: [email protected], interface: [::], port: 15674, protocol: http/web-stomp, purpose: STOMP over WebSockets
Node: [email protected], interface: [::], port: 25672, protocol: clustering, purpose: inter-node and CLI tool communication
Node: [email protected], interface: [::], port: 15672, protocol: http, purpose: HTTP API
Node: [email protected], interface: [::], port: 61613, protocol: stomp, purpose: STOMP
Node: [email protected], interface: [::], port: 1883, protocol: mqtt, purpose: MQTT
Node: [email protected], interface: [::], port: 5672, protocol: amqp, purpose: AMQP 0-9-1 and AMQP 1.0

Feature flags

Flag: drop_unroutable_metric, state: disabled
Flag: empty_basic_get_metric, state: disabled
Flag: implicit_default_bindings, state: enabled
Flag: maintenance_mode_status, state: enabled
Flag: quorum_queue, state: enabled
Flag: user_limits, state: enabled
Flag: virtual_host_metadata, state: enabled

[michaelklishin]

You can explicitly set cluster name to any value you like. By default it is "computed" from the name of the node itself:

-spec cluster_name() -> binary().

cluster_name() ->
    case rabbit_runtime_parameters:value_global(cluster_name) of 
        not_found -> cluster_name_default();
        Name -> Name
    end.

cluster_name_default() ->
    {ID, _} = parts(node()),
    FQDN = rabbit_net:hostname(),
    list_to_binary(atom_to_list(make({ID, FQDN}))).

so when a cluster is formed in parallel every node will have its own value. The name is not used by RabbitMQ except for some entities declared by Federation links, so I don't see any issue with it. There is no way to solve this without requiring a sequential node
startup (one by one).

[michaelklishin]

A couple more things are worth mentioning. You can explicitly set cluster ID via rabbitmq.conf which would make it available to every cluster member even very early on boot. Perhaps that's what you want here, instead of using CLI tools to set cluster name.

Also, nodes have an internal cluster ID that's not exposed to the user. That is used by Tanzu RabbitMQ (the commercial edition) and are not exposed to the user in any way. So when a node needs a unique cluster ID that's consistent, there is such mechanism and it cannot be changed by the users.

[mxmcarl]

OK， I‘ll set cluster ID via rabbitmq.conf.
When cluster ID is needed, internal ID is used instead of user set ID, right？

[michaelklishin]

The cluster ID you control is for your own convenience and as an extra identifier in e.g. federation link consumer tags.
AMQP 1.0 plugin uses it as a container ID value, too.

Unless your deployment tools depend on it one way or another, it's very unlikely that changing it will have any visible effects.