Upgrading to a Next Major Release

Upgrading to a Next Major Release

Upgrading a TripleO deployment to a next major release is done by first upgrading the undercloud, and then upgrading the overcloud.

Note that there are version specific caveats and notes which are pointed out as below:

Mitaka to Newton

mitaka to newton specific note

Note

You can use the “Limit Environment Specific Content” in the left hand nav bar to restrict content to the upgrade you are performing.

Note

Generic upgrade testing cannot cover all possible deployment configurations. Before performing the upgrade in production, test it in a matching staging environment, and create a backup of the production environment.

Upgrading the Undercloud

  1. Disable the old OpenStack release repositories and enable new release repositories on the undercloud:

Mitaka to Newton

export CURRENT_VERSION=mitaka
export NEW_VERSION=newton

Newton to Ocata

export CURRENT_VERSION=newton
export NEW_VERSION=ocata

Backup and disable current repos. Note that the repository files might be named differently depending on your installation:

mkdir /home/stack/REPOBACKUP
sudo mv /etc/yum.repos.d/delorean* /home/stack/REPOBACKUP/

Get and enable new repos for NEW_VERSION:

Stable Branch

Enable the appropriate repos for the desired release, as indicated below. Do not enable any other repos not explicitly marked for that release.

Newton

Enable latest RDO Newton Delorean repository for all packages

sudo curl -L -o /etc/yum.repos.d/delorean-newton.repo https://trunk.rdoproject.org/centos7-newton/current/delorean.repo

Enable the Newton Delorean Deps repository

sudo curl -L -o /etc/yum.repos.d/delorean-deps-newton.repo https://trunk.rdoproject.org/centos7-newton/delorean-deps.repo

Ceph

Enable the CentOS Storage SIG Ceph/Jewel repository if using Ceph

sudo yum -y install --enablerepo=extras centos-release-ceph-jewel
sudo sed -i -e 's%gpgcheck=.*%gpgcheck=0%' /etc/yum.repos.d/CentOS-Ceph-Jewel.repo

Ocata

Enable latest RDO Ocata Delorean repository for all packages

sudo curl -L -o /etc/yum.repos.d/delorean-ocata.repo https://trunk.rdoproject.org/centos7-ocata/current/delorean.repo

Enable the Ocata Delorean Deps repository

sudo curl -L -o /etc/yum.repos.d/delorean-deps-ocata.repo https://trunk.rdoproject.org/centos7-ocata/delorean-deps.repo

Ceph

Enable the CentOS Storage SIG Ceph/Jewel repository if using Ceph

sudo yum -y install --enablerepo=extras centos-release-ceph-jewel
sudo sed -i -e 's%gpgcheck=.*%gpgcheck=0%' /etc/yum.repos.d/CentOS-Ceph-Jewel.repo

Enable last known good RDO Trunk Delorean repository for core openstack packages

sudo curl -L -o /etc/yum.repos.d/delorean.repo https://trunk.rdoproject.org/centos7-master/current-passed-ci/delorean.repo

Enable latest RDO Trunk Delorean repository only for the TripleO packages

sudo curl -L -o /etc/yum.repos.d/delorean-current.repo https://trunk.rdoproject.org/centos7/current/delorean.repo
sudo sed -i 's/\[delorean\]/\[delorean-current\]/' /etc/yum.repos.d/delorean-current.repo
sudo /bin/bash -c "cat <<EOF>>/etc/yum.repos.d/delorean-current.repo

includepkgs=diskimage-builder,instack,instack-undercloud,os-apply-config,os-collect-config,os-net-config,os-refresh-config,python-tripleoclient,openstack-tripleo-common*,openstack-tripleo-heat-templates,openstack-tripleo-image-elements,openstack-tripleo,openstack-tripleo-puppet-elements,openstack-puppet-modules,openstack-tripleo-ui,puppet-*
EOF"

Enable the Delorean Deps repository

sudo curl -L -o /etc/yum.repos.d/delorean-deps.repo https://trunk.rdoproject.org/centos7/delorean-deps.repo

Ceph

Enable the CentOS Storage SIG Ceph/Jewel repository if using Ceph

sudo yum -y install --enablerepo=extras centos-release-ceph-jewel
sudo sed -i -e 's%gpgcheck=.*%gpgcheck=0%' /etc/yum.repos.d/CentOS-Ceph-Jewel.repo
  1. Run undercloud upgrade:

    Mitaka to Newton

    In the first release of instack-undercloud newton(5.0.0), the undercloud telemetry services are disabled by default. In order to maintain the telemetry services during the mitaka to newton upgrade the operator must explicitly enable them before running the undercloud upgrade. This is done by adding:

    enable_telemetry = true
    

    in the [DEFAULT] section of the undercloud.conf configuration file.

    If you are using any newer newton release, this option is switched back to enabled by default to make upgrade experience better. Hence, if you are using a later newton release you don’t need to explicitly enable this option.

    The following command will upgrade the undercloud:

    sudo systemctl stop openstack-*
    sudo systemctl stop neutron-*
    sudo systemctl stop httpd
    sudo yum -y update instack-undercloud openstack-puppet-modules openstack-tripleo-common python-tripleoclient
    openstack undercloud upgrade
    

    Once the undercloud upgrade is fully completed you may remove the older mysql backup folder /home/stack/mysql-backup

Note

You may wish to use time and capture the output to a file for any debug:

time openstack undercloud upgrade 2>&1 | tee undercloud_upgrade.log

Note

If you added custom OVS ports to the undercloud (e.g. in a virtual testing environment) you may need to re-add them at this point.

Upgrading the Overcloud to Ocata and beyond

As of the Ocata release, the upgrades workflow in tripleo has changed significantly to accommodate the operators’ new ability to deploy custom roles with the Newton release (see the Composable Service Upgrade spec for more info). The new workflow uses ansible upgrades tasks to define the upgrades workflow on a per-service level.

The operator starts the upgrade with a openstack overcloud deploy that includes the major-upgrade-composable-steps.yaml environment file as well as all other environment files used on the initial deployment. This will collect the ansible upgrade tasks for all roles, except those that have the disable_upgrade_deployment flag set True in roles_data.yaml. The tasks will be executed in a series of steps, for example (and not limited to): step 0 for validations or other pre-upgrade tasks, step 1 to stop the pacemaker cluster, step 2 to stop services, step 3 for package updates, step 4 for cluster startup, step 5 for any special case db syncs or post package update migrations.

After the ansible tasks have run the puppet configuration is also applied in the ‘normal’ manner we do on an initial deploy, to complete the upgrade and bring services back up.

For those roles with the disable_upgrade_deployment flag set True, the operator will upgrade the corresponding nodes with the upgrade-non-controller.sh. The operator uses that script to invoke the tripleo_upgrade_node.sh which is delivered during the major-upgrade-composable-steps that comes first, as described above.

  1. Run the major upgrade composable ansible steps

    This step will upgrade the nodes of all roles that do not explicitly set the disable_upgrade_deployment flag to True in the roles_data.yaml (this is an operator decision, and the current default is for the ‘Compute’ and’ ObjectStorage’ roles to have this set).

    The ansible upgrades tasks are collected from all service manifests and executed in a series of steps as described in the introduction above. Even before the invocation of these ansible tasks however, this upgrade step also delivers the tripleo_upgrade_node.sh and role specific puppet manifest to allow the operator to upgrade those nodes after this step has completed.

    Create an environment file with commands to switch OpenStack repositories to a new release. This will likely be the same commands that were used to switch repositories on the undercloud:

    cat > overcloud-repos.yaml <<EOF
    parameter_defaults:
      UpgradeInitCommand: |
        set -e
        # REPOSITORY SWITCH COMMANDS GO HERE
    EOF
    

    And run overcloud deploy, passing in full set of environment files plus major-upgrade-composable-steps.yaml and overcloud-repos.yaml:

    openstack overcloud deploy --templates \
        -e <full environment> \
        -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-composable-steps.yaml \
        -e overcloud-repos.yaml
    

Note

It is especially important to remember that you must include all environment files that were used to deploy the overcloud that you are about to upgrade.

Note

The first step of the ansible tasks is to validate that the deployment is in a good state before performing any other upgrade operations. Each service manifest in the tripleo-heat-templates includes a check that it is running and if any of those checks fail the upgrade will exit early at ansible step 0.

If you are re-running the upgrade after an initial failed attempt, you may need to disable these checks in order to allow the upgrade to proceed with services down. This is done with the SkipUpgradeConfigTags parameter to specify that tasks with the ‘validation’ tag should be skipped. You can include this in any of the environment files you are using:

SkipUpgradeConfigTags: [validation]
  1. Upgrade remaining nodes for roles with disable_upgrade_deployment: True

    It is expected that the operator will want to upgrade the roles that have the openstack-nova-compute and openstack-swift-object services deployed to allow for pre-upgrade migration of workfloads. For this reason the default Compute and ObjectStorage roles in the roles_data.yaml have the disable_upgrade_deployment set True.

    Note that unlike in previous releases, this operator driven upgrade step includes a full puppet configuration run as happens after the ansible steps on the roles those are executed on. The significance is that nodes are ‘fully’ upgraded after each step completes, rather than having to wait for the final converge step as has previously been the case.

    The tripleo_upgrade_node.sh script and puppet configuration are delivered to the nodes with disable_upgrade_deployment set True during the initial major upgrade composable steps in step 1 above.

    To upgrade remaining roles (at your convenience):

    upgrade-non-controller.sh --upgrade overcloud-compute-0
    
    for i in \$(seq 0 2); do
      upgrade-non-controller.sh --upgrade overcloud-objectstorage-\$i &"
    done
    
  2. Converge to unpin Nova RPC

    The final step is required to unpin Nova RPC version. Unlike in previous releases, for Ocata the puppet configuration has already been applied to nodes as part of each upgrades step, i.e. after the ansible tasks or when invoking the tripleo_upgrade_node.sh script to upgrade compute nodes. Thus the significance of this step is somewhat diminished compared to previously. However a re-application of puppet configuration across all nodes here will also serve as a sanity check and hopefully show any issues that an operator may have missed during any of the previous upgrade steps:

    openstack overcloud deploy --templates \
     -e <full environment> \
     -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-converge.yaml
    

Note

It is especially important to remember that you must include all environment files that were used to deploy the overcloud.

Upgrading the Overcloud to Newton and earlier

Note

The openstack overcloud deploy calls in upgrade steps below are non-blocking. Make sure that the overcloud is UPDATE_COMPLETE in openstack stack list and sudo pcs status on a controller reports everything running fine before proceeding to the next step.

Mitaka to Newton

Deliver the migration for ceilometer to run under httpd.

This is to deliver the migration for ceilometer to be run under httpd (apache) rather than eventlet as was the case before. To execute this step run overcloud deploy, passing in the full set of environment files plus major-upgrade-ceilometer-wsgi-mitaka-newton.yaml:

openstack overcloud deploy --templates \
    -e <full environment> \
    -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-ceilometer-wsgi-mitaka-newton.yaml
  1. Upgrade initialization

    The initialization step switches to new repositories on overcloud nodes, and it delivers upgrade scripts to nodes which are going to be upgraded one-by-one (this means non-controller nodes, except any stand-alone block storage nodes).

    Create an environment file with commands to switch OpenStack repositories to a new release. This will likely be the same commands that were used to switch repositories on the undercloud:

    cat > overcloud-repos.yaml <<EOF
    parameter_defaults:
      UpgradeInitCommand: |
        set -e
        # REPOSITORY SWITCH COMMANDS GO HERE
    EOF
    

    And run overcloud deploy, passing in full set of environment files plus major-upgrade-pacemaker-init.yaml and overcloud-repos.yaml:

    openstack overcloud deploy --templates \
        -e <full environment> \
        -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-pacemaker-init.yaml \
        -e overcloud-repos.yaml
    
  2. Object storage nodes upgrade

    If the deployment has any standalone object storage nodes, upgrade them one-by-one using the upgrade-non-controller.sh script on the undercloud node:

    upgrade-non-controller.sh --upgrade <nova-id of object storage node>
    

    This is ran before controller node upgrade because swift storage services should be upgraded before swift proxy services.

  3. Upgrade controller and block storage nodes

    Mitaka to Newton

    Explicitly disable sahara services if so desired: As discussed at bug1630247 sahara services are disabled by default in the Newton overcloud deployment. This special case is handled for the duration of the upgrade by defaulting to ‘keep sahara-*’.

    That is by default sahara services are restarted after the mitaka to newton upgrade of controller nodes and sahara config is re-applied during the final upgrade converge step.

    If an operator wishes to disable sahara services as part of the mitaka to newton upgrade they need to include the major-upgrade-remove-sahara.yaml environment file during the controller upgrade step as well as during the converge step later:

    openstack overcloud deploy --templates \
     -e <full environment> \
     -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-pacemaker.yaml
     -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-remove-sahara.yaml
    

    All controllers will be upgraded in sync in order to make services only talk to DB schema versions they expect. Services will be unavailable during this operation. Standalone block storage nodes are automatically upgraded in this step too, in sync with controllers, because block storage services don’t have a version pinning mechanism.

    Run the deploy command with major-upgrade-pacemaker.yaml:

    openstack overcloud deploy --templates \
        -e <full environment> \
        -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-pacemaker.yaml
    

    Services of the compute component on the controller nodes are now pinned to communicate like the older release, ensuring that they can talk to the compute nodes which haven’t been upgraded yet.

    Note

    If this step fails, it may leave the pacemaker cluster stopped (together with all OpenStack services on the controller nodes). The root cause and restoration procedure may vary, but in simple cases the pacemaker cluster can be started by logging into one of the controllers and running sudo pcs cluster start –all.

  4. Upgrade ceph storage nodes

    If the deployment has any ceph storage nodes, upgrade them one-by-one using the upgrade-non-controller.sh script on the undercloud node:

    upgrade-non-controller.sh --upgrade <nova-id of ceph storage node>
    
  5. Upgrade compute nodes

    Upgrade compute nodes one-by-one using the upgrade-non-controller.sh script on the undercloud node:

    upgrade-non-controller.sh --upgrade <nova-id of compute node>
    
  6. Apply configuration from upgraded tripleo-heat-templates

    Mitaka to Newton

    Explicitly disable sahara services if so desired: As discussed at bug1630247 sahara services are disabled by default in the Newton overcloud deployment. This special case is handled for the duration of the upgrade by defaulting to ‘keep sahara-*’.

    That is by default sahara services are restarted after the mitaka to newton upgrade of controller nodes and sahara config is re-applied during the final upgrade converge step.

    If an operator wishes to disable sahara services as part of the mitaka to newton upgrade they need to include the major-upgrade-remove-sahara.yaml environment file during the controller upgrade earlier and converge step here:

    openstack overcloud deploy --templates \
     -e <full environment> \
     -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-pacemaker-converge.yaml
     -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-remove-sahara.yaml
    

    This step unpins compute services communication (upgrade level) on controller and compute nodes, and it triggers configuration management tooling to converge the overcloud configuration according to the new release of tripleo-heat-templates.

    Make sure that all overcloud nodes have been upgraded to the new release, and then run the deploy command with major-upgrade-pacemaker-converge.yaml:

    openstack overcloud deploy --templates \
        -e <full environment> \
        -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-pacemaker-converge.yaml
    

Mitaka to Newton

Deliver the data migration for aodh.

This is to deliver the data migration for aodh. In Newton, aodh uses its own mysql backend. This step migrates all the existing alarm data from mongodb to the new mysql backend. To execute this step run overcloud deploy, passing in the full set of environment files plus major-upgrade-aodh-migration.yaml:

openstack overcloud deploy --templates \
    -e <full environment> \
    -e /usr/share/openstack-tripleo-heat-templates/environments/major-upgrade-aodh-migration.yaml
Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.