VMware Tanzu RabbitMQ [VMs] Release Notes

Note: Pivotal Platform is now part of VMware Tanzu. In v1.19 and later, RabbitMQ for Pivotal Platform is named VMware Tanzu RabbitMQ [VMs].

For product versions and upgrade paths, see Upgrade Planner.

v1.19.5

Release Date: December 14, 2020

Features

New features and changes in this release:

• Option to limit the TLS versions enabled for on-demand instances: VMware Tanzu RabbitMQ v1.21.0 supports TLS v1.0, v1.1, and v1.2 (default). Additionally, you can now select the TLS versions to enable in RabbitMQ. This disables the other TLS versions.
  To configure TLS, follow the procedure in Configure Security.
  For more information about disabling TLS versions, see the RabbitMQ documentation.

• Option to add VM extensions to on-demand instances: You can provide a list of VM extensions to add to on-demand service instances. This feature enables you to exclude VMs hosting on-demand service instances from being dynamically provisioned by an underlying IaaS, among other capabilities. For more information, see Configure Global Settings in Installing and Configuring the On-Demand Service.

• Improved internal security for service-gateway: For delete operations, the on-demand service broker now communicates with routing components over mutual TLS.

Known Issues

This release has the following issue:

• Changing the Erlang cookie can cause failed deployments: Changing the Erlang cookie value requires cluster downtime and can result in failed deployments. For more information, see Changing the Erlang Cookie Value Known Issue.

Compatibility

The following components are compatible with this release:

^* For more information, see the v3.8.9 GitHub documentation.

v1.19.4

Release Date: September 23, 2020

Features

New features and changes in this release:

• Erlang: Upgraded to v23.0.3.

Resolved Issues

This release has the following fix:

• The on-demand adapter cleanly exits the drain script when the credentials are not present in UAA: The drain script now has the correct output for success (0) and failure (non-zero).

Known Issues

This release has the following issue:

• Changing the Erlang cookie can cause failed deployments: Changing the Erlang cookie value requires cluster downtime and can result in failed deployments. For more information, see Changing the Erlang Cookie Value Known Issue.

Compatibility

The following components are compatible with this release:

^* For more information, see the v3.8.8 GitHub documentation.

v1.19.3

Release Date: August 5, 2020

Resolved Issues

This release has the following fixes:

• Configure a custom apps domain to ensure smoke tests work: Previously, the smoke test picked up the default Cloud Foundry domain, which could cause smoke tests to fail when the RabbitMQ service instance domain for on-demand services could not access the smoke tests domain.
  Operators can now configure the smoke tests domain to ensure that the service instance is accessible during smoke tests.
  For more information, see Configure Global Settings.

• The pre-stop script no longer fails with a timeout error. Previously, the pre-stop script failed with the error: Timed out waiting for mirror queue critical node to sync after 3600 seconds. To prevent this, the following checks in the pre-stop script are disabled by default:

  • rabbitmq-diagnostics check_if_node_is_mirror_sync_critical
  • rabbitmq-diagnostics check_if_node_is_quorum_critical

  To enable these checks, select the Wait for Queue Synchronization checkbox in the configuration for on-demands plans. For more information about this setting, see Configure the Service Plan.

• Downstream federated queues consuming messages from an upstream queue are no longer duplicated after upgrading. This is because the RabbitMQ cluster name defaults to rabbit@localhost instead of its service instance GUID. To set the cluster name to its service instance GUID, select the Use Service Instance ID as cluster name checkbox in Global Settings for On-Demand Plans. For more information about this setting, see Configure Global Settings.

Known Issues

This release has the following issues:

• Changing the Erlang cookie can cause failed deployments: Changing the Erlang cookie value requires cluster downtime and can result in failed deployments. For more information, see Changing the Erlang Cookie Value Known Issue.

• The on-demand adapter does not cleanly exit the drain script when the credentials are not present in UAA. The drain script is expected to output an integer to determine success (0) or failure (non-zero). In some cases, the drain script outputs errors even though the operation has succeeded. This causes a failure in the drain operation. This is fixed in v1.19.4.

Compatibility

The following components are compatible with this release:

^* For more information, see the v3.8.5 GitHub documentation.

v1.19.1

Release Date: June 10, 2020

Resolved Issues

This release has the following fixes:

• Upgrading a replicated tile no longer fails with the error this migration in a replicated tile caused the error "TypeError: Cannot read property 'value' of undefined".
• BOSH deployments no longer fail when the RabbitMQ admin password for the pre-provisioned cluster contains a special character.
• To prevent rabbitmqctl node_health_check timing out after 70 seconds, RabbitMQ has less intrusive health checks. For more information, see the RabbitMQ documentation.

Known Issues

This release has the following issues:

• Changing the Erlang cookie can cause failed deployments: Changing the Erlang cookie value requires cluster downtime and can result in failed deployments. For more information, see Changing the Erlang Cookie Value Known Issue.

• The default setting for mirrored queues can cause the pre-stop script to timeout: In production, if you use mirrored queues with the default policy setting ha-sync-mode: manual and if you have not manually synced your queues, the pre-stop script can fail with the following error: "Timed out waiting for mirror queue critical node to sync after 3600 seconds". To manually sync your queues, run rabbitmqctl sync_queue. To set ha-sync-mode to automatic instead, see Setting or Changing the Policy.

• After upgrading, the downstream federated queues that consume messages from an upstream queue are duplicated: Federated queues are duplicated because VMware Tanzu RabbitMQ recreates them with a new cluster name. This issue occurs the first time you run the upgrade-all-service-instances errand after upgrading to VMware Tanzu RabbitMQ v1.18.5 or later, and does not reoccur if you upgrade to subsequent versions.

• Smoke tests fail if the RabbitMQ service instance domain cannot access the test domain: Smoke tests pick up the default Cloud Foundry domain, which can cause them to fail if the RabbitMQ service instance domain for on-demand services cannot access the smoke tests domain.

• The on-demand adapter does not cleanly exit the drain script when the credentials are not present in UAA. The drain script is expected to output an integer to determine success (0) or failure (non-zero). In some cases, the drain script outputs errors even though the operation has succeeded. This causes a failure in the drain operation. This is fixed in v1.19.4.

Compatibility

The following components are compatible with this release:

^* For more information, see the v3.8.3 GitHub documentation.

v1.19.0

Release Date: May 15, 2020

Features

New features in this release:

• RabbitMQ Prometheus plugin support: This plugin collects RabbitMQ node metrics that you can use for monitoring. To enable this feature, you must open port 15692.
  
  For information about the RabbitMQ Prometheus plugin, see the RabbitMQ documentation.
  For more information about enabling this feature, see Required Networking Rules for On-Demand Services.

• Service-Gateway Improvement: Operators can configure an on-demand plan that enables service instances to connect to components outside the TAS for VMs foundation through MQTT and STOMP. For more information, see Enabling Service-Gateway Access.

• Filter metrics by cluster names: The RabbitMQ cluster name now matches its service instance GUID, enabling operators to filter metrics by cluster names.

Resolved Issues

This release has the following fixes:

• Quorum queue downtime during rolling upgrades: When used in production, quorum queues no longer experience downtime during a rolling upgrade. This is because VMware Tanzu RabbitMQ waits for quorum and mirror-sync critical nodes to sync before upgrades.

Known Issues

This release has the following issues:

• Changing the Erlang cookie can cause failed deployments: Changing the Erlang cookie value requires cluster downtime and can result in failed deployments. For more information, see Changing the Erlang Cookie Value Known Issue.

• Upgrading a replicated tile fails with the error this migration in a replicated tile caused the error "TypeError: Cannot read property 'value' of undefined"

• In production use cases, rabbitmqctl node_health_check times out after 70 seconds.

• The default setting for mirrored queues can cause the pre-stop script to timeout: In production, if you use mirrored queues with the default policy setting ha-sync-mode: manual and if you have not manually synced your queues, the pre-stop script can fail with the following error: "Timed out waiting for mirror queue critical node to sync after 3600 seconds". To manually sync your queues, run rabbitmqctl sync_queue. To set ha-sync-mode to automatic instead, see Setting or Changing the Policy.

• After upgrading, the downstream federated queues that consume messages from an upstream queue are duplicated: Federated queues are duplicated because VMware Tanzu RabbitMQ recreates them with a new cluster name. This issue occurs the first time you run the upgrade-all-service-instances errand after upgrading to VMware Tanzu RabbitMQ v1.18.5 or later, and does not reoccur if you upgrade to subsequent versions.

• Smoke tests fail if the RabbitMQ service instance domain cannot access the test domain: Smoke tests pick up the default Cloud Foundry domain, which can cause them to fail if the RabbitMQ service instance domain for on-demand services cannot access the smoke tests domain.

• The on-demand adapter does not cleanly exit the drain script when the credentials are not present in UAA. The drain script is expected to output an integer to determine success (0) or failure (non-zero). In some cases, the drain script outputs errors even though the operation has succeeded. This causes a failure in the drain operation. This is fixed in v1.19.4.

Compatibility

The following components are compatible with this release:

^* For more information, see the v3.8.3 GitHub documentation.

View Release Notes for Another Version

To view the release notes for another product version, select the version from the dropdown at the top of this page.