Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

 
ON THIS PAGE
 

Upgrade Procedure for Ubuntu-based Contrail 4.1.3 to Contrail 4.1.4 Using Juju with Netronome SmartNIC

This section presents the steps to upgrade from an Ubuntu-16.04-based Contrail deployment from Contrail version 4.1.3 to Contrail version 4.1.4 using Juju charms.

Prerequisites

These instructions assume that these requirements for installing Contrail Release 4.1.3 are already present in your environment:

  • MaaS Server—MaaS version 2.3 is installed on this server. This procedure was tested using MaaS version 2.3.5.

  • Juju Controller—Juju version 2.3 is installed, and the Juju controller is bootstrapped on a VM or a bare metal server. This procedure was tested using Juju version 2.3.7-xenial-amd64.

  • A repository to get Netronome, patched Openstack packages, and Contrail vRouter packages is operational.

  • A Contrail Controller using Ubuntu 16.04 xenial is operational.

  • A Contrail cluster with one or more compute nodes using Agilio SmartNICs.

Acquire the Software

To acquire the Contrail 4.1.4 software images to perform this procedure:

  1. Go to the Juniper Networks Support site for Contrail.

  2. Select OS as Contrail and Version as 4.1.4 from the drop-down menus.

  3. Download the contrail-cloud-docker_4.1.4.0-63-ocata_xenial.tgz file.

  4. Extract the following images from the contrail-cloud-docker_4.1.4.0-63-ocata_xenial.tgz file:

    • Contrail Analytics package: contrail-analytics-ubuntu16.04-4.1.4.0-63.tar.gz.

    • Contrail Analytics Database package: contrail-analyticsdb-ubuntu16.04-4.1.4.0-63.tar.gz .

    • Contrail Controller package: contrail-controller-ubuntu16.04-4.1.4.0-63.tar.gz

    The images need to be extracted because the Contrail Analytics, Contrail Analytics Database, and Contrail Controller packages must be upgraded individually to perform this upgrade.

Attach Contrail Packages using Juju

The Contrail Controller, Contrail Analytics, and Contrail Analytics DB packages need to be attached using Juju to perform this upgrade.

To attach these software packages into Juju:

  1. Attach the Contrail Controller, Contrail Analytics, & Contrail Analytics DB packages into Juju:

  2. Check status of the software image attachments into Juju using the juju status command.

    Wait for the juju status command output to indicate that the upgrade is successful. The output in the juju status should indicate that all processes are Active and all machine states are started.

Upgrade the Contrail Clusters

This section provides the steps to update the Contrail clusters for this upgrade.

It includes the following sections:

Upgrade the Contrail Controllers

The Contrail controllers must be upgraded one by one to complete this procedure.

To upgrade the Contrail controllers:

  1. SSH into the Contrail controller server and decommission the Contrail controller from the Cassandra cluster:

  2. Remove the Contrail Controller container:

  3. Update the hooks to the Contrail Controller from the Juju Controller:

  4. Wait for the Contrail status for all packages on the upgrading node to change to active. This step can take up to 10 minutes.

    Enter the contrail-status command to check status. All packages in the Contrail Control section of the output must move to the active state before proceeding.

  5. Check Juju status by entering the juju status command.

    All Contrail components in this output should be in the active state.

  6. After each controller update, check the controllers to make sure the databases are consistent across all controllers:

    • Enter the nodetool describecluster command. Confirm that the schema version output is identical on all 3 controllers.

    • Enter the echo stat | nc localhost 2181 command. The node count output should be identical on all 3 controllers.

    • Ensure that the contrail-status output is active for all components in all 3 controllers.

    If your upgrade is not successful after 15 minutes, retry steps 1 through 5.

    If you need to decommission a node that is not upgrading successfully, use the nodetool removenode node-ID command.

  7. Repeat steps 1 through 6 for all other Contrail controller nodes.

Upgrade Contrail Analytics Nodes

To upgrade the Contrail Analytics nodes:

  1. SSH into the first Contrail Analytics node and remove the Contrail Analytics container:

  2. Confirm Juju status using the juju status command.

    The output in the juju status should indicate that all processes are Active and all machine states are started.

  3. From the MaaS server, update hooks to the Contrail Analytics controller:

  4. Wait for the Contrail status for all packages on the upgrading node to change to active. This step can take up to 10 minutes.

    Enter the contrail-status command to check status. All packages in the Contrail Analytics section of the output must move to the active state before proceeding.

  5. Repeat steps 1 through 4 for all other Contrail Analytics nodes.

Upgrade Analytics Database Nodes

To upgrade the Contrail Analytics database nodes:

  1. SSH into a Contrail analytics database server and decommission the node from the Cassandra cluster:

  2. Remove the AnalyticsDB container:

  3. From the Juju controller, update the hooks to the Contrail Analytics DB controller:

  4. Wait for the Contrail status for all packages on the upgrading node to change to active. This step can take up to 10 minutes.

    Enter the contrail-status command to check status. All packages in the Contrail Database section of the output must move to the active state before proceeding.

  5. Check Juju status by entering the juju status command.

    All Contrail components in this output should be in the active state.

  6. After each analytics database node update, check the nodes to ensure the databases are consistent inside the contrail analytics database containers:

    • Enter the nodetool describecluster command. Confirm that the schema version output is identical on all 3 nodes.

    • Enter the echo stat | nc localhost 2181 command. The node count output should be identical on all 3 nodes.

    • Ensure that the contrail-status output is active for all components in all 3 contrail analytics db nodes.

    If your upgrade is not successful after 15 minutes, retry steps 1 through 5.

    If you need to decommission a node that is not upgrading successfully, use the nodetool removenode node-ID command.

  7. Repeat steps 1 through 6 for all other Contrail Analytics database nodes.

Updating the Neutron Plugin and the vRouter Agent

The process for updating the neutron plugin and the vRouter agent is different for compute nodes than it is for other nodes.

This section covers both procedures and includes these sections:

Updating the Neutron Plugin and the vRouter Agent on Non-Compute Nodes

Use this procedure to update the Neutron Plugin and the vRouter agent on all non-compute nodes in your environment:

Note:

This procedure assumes that the APT Get repository was created during the previous installation, and that the latest Contrail packages can be placed into the repository.

  1. SSH into the Neutron API plugin unit.

  2. From the Neutron API plugin unit, get the latest APT Get update:

  3. Upgrade APT GET:

    Note:

    This step shows how to upgrade APT get for all packages. You can also manually update the neutron-plugin-contrail and python-contrail packages to complete this step, if you’d rather not perform the complete upgrade. This procedure does not provide the steps to manually update these packages.

  4. Restart the Neutron service:

Updating the Neutron Plugin and the vRouter Agent on Compute Nodes

Use this procedure to update the Neutron Plugin and the vRouter agent on all compute devices in your environment:

Note:

This procedure assumes that the APT Get repository was created during the previous installation, and that the latest Contrail packages can be placed into the repository.

  1. SSH into the Neutron API plugin unit.

  2. From the Neutron API plugin unit, get the latest APT Get update:

  3. Upgrade APT GET:

    Note:

    This step shows how to upgrade APT get for all packages. You can also manually update the following packages to complete this step:

    • contrail-lib

    • contrail-nodemgr

    • contrail-setup

    • contrail-utils

    • contrail-vrouter-agent

    • contrail-vrouter-common

    • contrail-vrouter-dkms

    • contrail-vrouter-init

    • contrail-vrouter-utils

    • python-contrail

    • python-contrail-vrouter-api

    • python-opencontrail-vrouter-netns

    This procedure does not provide the steps to manually update these packages.

  4. Upgrade the vRouter agent and, if using Netronome SmartNICs, the netronome plugin.

    • If you are performing this procedure on a compute node without a Netronome SmartNIC:

      Note:

      The network connection over the vhost is down while this procedure is performed. Traffic will be lost.

      1. Stop the Contrail vRouter agent:

      2. Remove the Contrail vRouter module:

      3. Insert the vRouter module:

      4. Activate the vhost:

      5. Restart the Contrail vRouter agent:

    • If you are performing this procedure on a compute node with a Netronome SmartNIC:

      Note:

      The network connection over the vhost is down while this procedure is performed. Traffic will be lost.

      1. Stop the Contrail vRouter agent:

      2. Stop the Virtio forwarder module:

      3. Stop the vRouter control module:

      4. Restart the Virtio forwarder module:

      5. Restart the Contrail vRouter agent:

      6. Activate the vhost:

      7. Restart the Contrail vRouter agent:

  5. Verify Contrail status:

    All packages in the Contrail vRouter section of the output should be in the active state. This step can take several minutes.

Related Documentation