RabbitMQ

RabbitMQ is a message broker written in Erlang. It is currently the default provider of message queues in Kolla Ansible deployments.

TLS encryption

There are a number of channels to consider when securing RabbitMQ communication. Kolla Ansible currently supports TLS encryption of the following:

  • client-server traffic, typically between OpenStack services using the oslo.messaging library and RabbitMQ

  • RabbitMQ Management API and UI (frontend connection to HAProxy only)

Encryption of the following channels is not currently supported:

  • RabbitMQ cluster traffic between RabbitMQ server nodes

  • RabbitMQ CLI communication with RabbitMQ server nodes

Client-server

Encryption of client-server traffic is enabled by setting rabbitmq_enable_tls to true. Additionally, certificates and keys must be available in the following paths (in priority order):

Certificates:

  • "{{ kolla_certificates_dir }}/{{ inventory_hostname }}/rabbitmq-cert.pem"

  • "{{ kolla_certificates_dir }}/{{ inventory_hostname }}-cert.pem"

  • "{{ kolla_certificates_dir }}/rabbitmq-cert.pem"

Keys:

  • "{{ kolla_certificates_dir }}/{{ inventory_hostname }}/rabbitmq-key.pem"

  • "{{ kolla_certificates_dir }}/{{ inventory_hostname }}-key.pem"

  • "{{ kolla_certificates_dir }}/rabbitmq-key.pem"

The default for kolla_certificates_dir is /etc/kolla/certificates.

The certificates must be valid for the IP address of the host running RabbitMQ on the API network.

Additional TLS configuration options may be passed to RabbitMQ via rabbitmq_tls_options. This should be a dict, and the keys will be prefixed with ssl_options.. For example:

rabbitmq_tls_options:
  ciphers.1: ECDHE-ECDSA-AES256-GCM-SHA384
  ciphers.2: ECDHE-RSA-AES256-GCM-SHA384
  ciphers.3: ECDHE-ECDSA-AES256-SHA384
  honor_cipher_order: true
  honor_ecc_order: true

Details on configuration of RabbitMQ for TLS can be found in the RabbitMQ documentation.

When om_rabbitmq_enable_tls is true (it defaults to the value of rabbitmq_enable_tls), applicable OpenStack services will be configured to use oslo.messaging with TLS enabled. The CA certificate is configured via om_rabbitmq_cacert (it defaults to rabbitmq_cacert, which points to the system’s trusted CA certificate bundle for TLS). Note that there is currently no support for using client certificates.

For testing purposes, Kolla Ansible provides the kolla-ansible certificates command, which will generate self-signed certificates for RabbitMQ if rabbitmq_enable_tls is true.

Management API and UI

The management API and UI are accessed via HAProxy, exposed only on the internal VIP. As such, traffic to this endpoint is encrypted when kolla_enable_tls_internal is true. See TLS Configuration.

Passing arguments to RabbitMQ server’s Erlang VM

Erlang programs run in an Erlang VM (virtual machine) and use the Erlang runtime. The Erlang VM can be configured.

Kolla Ansible makes it possible to pass arguments to the Erlang VM via the usage of the rabbitmq_server_additional_erl_args variable. The contents of it are appended to the RABBITMQ_SERVER_ADDITIONAL_ERL_ARGS environment variable which is passed to the RabbitMQ server startup script. Kolla Ansible already configures RabbitMQ server for IPv6 (if necessary). Any argument can be passed there as documented in https://www.rabbitmq.com/runtime.html

The default value for rabbitmq_server_additional_erl_args is +S 2:2 +sbwt none +sbwtdcpu none +sbwtdio none.

By default RabbitMQ starts N schedulers where N is the number of CPU cores, including hyper-threaded cores. This is fine when you assume all CPUs are dedicated to RabbitMQ. Its not a good idea in a typical Kolla Ansible setup. Here we go for two scheduler threads (+S 2:2). More details can be found here: https://www.rabbitmq.com/runtime.html#scheduling and here: https://erlang.org/doc/man/erl.html#emulator-flags

The +sbwt none +sbwtdcpu none +sbwtdio none arguments prevent busy waiting of the scheduler, for more details see: https://www.rabbitmq.com/runtime.html#busy-waiting.

High Availability

With the release of RabbitMQ 4.0, all queues are highly available as they are configured to be quorum queues by default. RabbitMQ also offer queues called streams, which can be used to replace “fanout” queues with a more performant alternative. This is enabled by default, but can be disabled by setting om_enable_rabbitmq_stream_fanout: false. When changing queues to a different type, the follow procedure will be needed.

Warning

Since the default changed to have all queues be of durable type in the Epoxy release, following procedure is required to be carried out before any upgrade to Epoxy.

  1. Generate the new config for all services. After this, make sure not to restart any containers until after the RabbitMQ state has been reset.

    kolla-ansible genconfig

  2. Stop all OpenStack services which use RabbitMQ, so that they will not attempt to recreate any queues yet.

    kolla-ansible stop --tags <service-tags>

  3. Reconfigure RabbitMQ if you were previously using om_enable_rabbitmq_high_availability.

    kolla-ansible reconfigure --tags rabbitmq

  4. Reset the state on each RabbitMQ, to remove the old transient queues and exchanges.

    kolla-ansible rabbitmq-reset-state

  5. Start the OpenStack services again, at which point they will recreate the appropriate queues as durable.

    kolla-ansible deploy --tags <service-tags>

Upgrading RabbitMQ

RabbitMQ upgrades in Kolla Ansible are typically restricted to a single minor version increment at a time (e.g., from 4.0.x to 4.1.x). This is a safety measure to ensure that RabbitMQ’s internal data migrations and feature flags are processed correctly.

In some cases, specific multi-version upgrade paths are supported (for example, jumping from 3.13 directly to 4.2). These allowed paths are defined using the rabbitmq_allowed_upgrades variable in the RabbitMQ role defaults.

Operators can customize or extend these allowed upgrade paths by overriding this variable in globals.yml.

rabbitmq_allowed_upgrades:
  "3.13":
    - "4.0"
    - "4.1"
    - "4.2"
  "4.0":
    - "4.1"
    - "4.2"

If an invalid upgrade path is detected, the deployment will fail with a descriptive error message during the rabbitmq-version-check task, suggesting the next appropriate intermediate version.

Handling Stream Replicas

RabbitMQ streams are expected to be replicated across the nodes in the cluster. However, RabbitMQ itself will only create replicas of a stream when the stream is initially declared. This means that any streams declared when a RabbitMQ node is out of service must be explicitly managed by an operator. RabbitMQ documents how to manage stream replicas here: https://www.rabbitmq.com/docs/streams#member-management

An example script to create any missing stream replicas can be found under kolla-ansible/contrib/ops/rabbitmq/rabbitmq-repair-stream-replicas.sh. This should be executed from a host running the RabbitMQ container. Currently, membership changes for streams is not entirely safe, so this script should only be used when the RabbitMQ cluster is in a known healthy state.