Package and Container Updates

This section describes the Release Train process of creating new package repository snapshots and updating the container images to use these new packages. It includes instructions on the recommended way to test these updates, using the multinode test environment.

Preparations

  1. Before building images, you should check for any outstanding PRs into the earliest supported release. Below are the links for the 2023.1 (Antelope) branches.

  1. Follow the workflows documented here. Sync the package repositories. Then, for each release:

  • Make sure the nightly sync of package repositories has succeeded.

  • Update the Kayobe package repository versions.

  • Build and push Kolla container images.

  • Open a draft PR into stackhpc-kayobe-config with the updated container image tags.

  • Test.

  • Review the PR. Once approved, promote the container images before then merging the PR.

The rest of this document describes the final “test” stage.

Testing

The following steps describe the process to test the new package and container repositories. See the subsections beneath for further explanations.

  1. Build two multinode environments for OVS and OVN, both on the earliest supported release.

  2. Run tests on current package versions as a baseline.

  3. Upgrade host packages.

  4. Upgrade containers.

  5. Run tests again with the new packages.

  6. Request reviews for your proposed PR to bring in the new packages.

  7. Promote these packages before merging the PR, but after the CI checks have passed.

  8. Upgrade OpenStack to the next release.

  9. Repeat steps 2 and 4-7. (Step 3 is skipped as the host packages will be shared across these releases.)

  10. Repeat 8 and 9 for any further releases.

Creating the multinode environments

There is a comprehensive guide to setting up a multinode environment with Terraform, found here: https://github.com/stackhpc/terraform-kayobe-multinode. There are some things to note:

  • OVN is enabled by default, you should override it under etc/kayobe/environments/ci-multinode/kolla.yml kolla_enable_ovn: false for the OVS multinode environment.

  • Remember to set different vxlan_vnis for each.

  • Before starting any tests, run dnf distro-sync on each host to ensure you are using the same snapshots as in the release train. You can do this using the following commands:

    kayobe seed host command run -b --command "dnf distro-sync"
kayobe overcloud host command run -b --command "dnf distro-sync"

  • This may have installed a new kernel version. If so, you will need to reboot the overcloud hosts. You can check the installed kernels and the currently running kernel with the following commands. If the latest listed version is not running, you will need to reboot.

kayobe seed host command run -b --show-output --command "dnf list installed kernel"
kayobe seed host command run -b --show-output --command "uname -a"

kayobe playbook run --limit seed,overcloud $KAYOBE_CONFIG_PATH/ansible/reboot.yml

  • The tempest tests run automatically at the end of deploy-openstack.sh. If you have the time, it is worth fixing any failing tests you can so that there is greater coverage for the package updates. (Also remember to propose these fixes in the relevant repos where applicable.)

Upgrading host packages

Checkout the new kayobe-config branch (from the draft PR):

cd $KAYOBE_CONFIG_ROOT
git fetch
git checkout <branch-name>

For Rocky Linux 9, bump the snapshot versions in /etc/yum/repos.d with:

kayobe overcloud host configure -t dnf

Install new packages:

kayobe overcloud host package update --packages "*"

Perform a rolling reboot of hosts:

export ANSIBLE_SERIAL=1
kayobe playbook run $KAYOBE_CONFIG_PATH/ansible/reboot.yml --limit controllers
kayobe playbook run $KAYOBE_CONFIG_PATH/ansible/reboot.yml --limit compute[0]

# Test live migration
openstack server create --image cirros --flavor m1.tiny --network external --hypervisor-hostname antelope-pkg-refresh-ovs-compute-02.novalocal --os-compute-api-version 2.74 server1
openstack server migrate --live-migration server1
watch openstack server show server1

kayobe playbook run $KAYOBE_CONFIG_PATH/ansible/reboot.yml --limit compute[1]

# Try and migrate back
openstack server migrate --live-migration server1
watch openstack server show server1

Upgrading containers within a release

Checkout the new kayobe-config branch (from the draft PR):

cd $KAYOBE_CONFIG_ROOT
git fetch
git checkout <branch-name>

Deploy the services, once the new tags are set in the kayobe_config:

kayobe overcloud service deploy

Upgrading OpenStack to the next release in a multinode environment

Warning

This guide was written for the Wallaby release and has not been validated for Antelope. Proceed with caution.

As this is not a full production system, only a reduced number of steps need to be followed to upgrade to a new release. Below describes these steps, with stackhpc/wallaby as the starting branch:

cd <base_path>/src/kayobe-config/
git fetch
git checkout -b xena_uber_merge
git merge origin/stackhpc/xena

source <base_path>/venvs/kayobe/bin/activate
cd <base_patch>/src/kayobe
git checkout stackhpc/xena
git fetch
pip install -U ~/src/kayobe

kayobe control host upgrade
kayobe overcloud host upgrade

kayobe overcloud container image pull

---Optional
kayobe overcloud service configuration save --output-dir config/wallaby
kayobe overcloud service configuration generate --node-config-dir /tmp/kolla-xena-config
kayobe overcloud service configuration save --output-dir config/xena --node-config-dir /tmp/kolla-xena-config
kayobe overcloud host command run --command 'rm -rf /tmp/kolla-xena-config' --become
# Check the diff between the old and new configs
---

kayobe overcloud service upgrade

Tests

Tempest

Run tempest, you can then perform the other tests while it runs. Once complete, check if any tests are failing.

As of February 2023, only one test was expected to fail. This may no longer be the case, so any additional failures are worth exploring.

tempest.scenario.test_network_basic_ops.TestNetworkBasicOps.test_port_security_macspoofing_port

Poke around horizon

Perform some basic operations such as spawning VMs or attaching/detaching volumes and check that each page works correctly.

Monitoring

Check for any ERROR log messages in Kibana.

Check that the Grafana dashboards are all populated with data.

Check that there are no active alerts.

Check that there are no flapping alerts.

Octavia (OVN only)

You will need to add an Ubuntu image and create a keypair.

wget http://cloud-images.ubuntu.com/focal/current/focal-server-cloudimg-amd64.img

openstack image create \
    --progress \
    --container-format bare \
    --disk-format qcow2 \
    --file focal-server-cloudimg-amd64.img \
    Ubuntu-20.04

openstack keypair create --private-key ~/.ssh/os-admin os-admin

Then run Octavia test script:

https://gist.github.com/MoteHue/ee5990bddea0677f54d8bb93d307aa71#file-octavia_test-sh

Attempt to build OFED against the latest kernel

Note that this only needs to be performed once.

kayobe seed host configure -t dnf -kt none
kayobe seed host package update --packages "*"

Then run the OFED test script:

https://gist.github.com/cityofships/b4883ee19f75d14534f04115892b8465