Transitioning to Containerized Services

Each of the main Red Hat OpenStack Platform services run in containers. This provides a method of keep each service within its own isolated namespace separated from the host. This means:

Red Hat OpenStack Platform uses a set of containers built and managed with the kolla toolset.

The overcloud configuration requires initial registry configuration to determine where to obtain images and how to store them. Complete the following steps to generate and customize an environment file that you can use to prepare your container images.

The default file for preparing your containers (containers-prepare-parameter.yaml) contains the ContainerImagePrepare heat parameter. This parameter defines a list of strategies for preparing a set of images:

parameter_defaults:
  ContainerImagePrepare:
  - (strategy one)
  - (strategy two)
  - (strategy three)
  ...

Each strategy accepts a set of sub-parameters that defines which images to use and what to do with the images. The following table contains information about the sub-parameters you can use with each ContainerImagePrepare strategy:

The set parameter accepts a set of key: value definitions:

The ContainerImageRegistryCredentials parameter maps a container registry to a username and password to authenticate to that registry.

If a container registry requires a username and password, you can use ContainerImageRegistryCredentials to include credentials with the following syntax:

  ContainerImagePrepare:
  - push_destination: true
    set:
      namespace: registry.redhat.io/...
      ...
  ContainerImageRegistryCredentials:
    registry.redhat.io:
      my_username: my_password

In the example, replace my_username and my_password with your authentication credentials. Instead of using your individual user credentials, Red Hat recommends creating a registry service account and using those credentials to access registry.redhat.io content. For more information, see "Red Hat Container Registry Authentication".

The ContainerImageRegistryLogin parameter is used to control the registry login on the systems being deployed. This must be set to true if push_destination is set to false or not used.

  ContainerImagePrepare:
  - set:
      namespace: registry.redhat.io/...
      ...
  ContainerImageRegistryCredentials:
    registry.redhat.io:
      my_username: my_password
  ContainerImageRegistryLogin: true

The value of the ContainerImagePrepare parameter is a YAML list. This means that you can specify multiple entries. The following example demonstrates two entries where director uses the latest version of all images except for the nova-api image, which uses the version tagged with 16.0-44:

ContainerImagePrepare:
- tag_from_label: "{version}-{release}"
  push_destination: true
  excludes:
  - nova-api
  set:
    namespace: registry.redhat.io/rhosp-rhel8
    name_prefix: openstack-
    name_suffix: ''
    tag: 16.0
- push_destination: true
  includes:
  - nova-api
  set:
    namespace: registry.redhat.io/rhosp-rhel8
    tag: 16.0-44

The includes and excludes parameters use regular expressions to control image filtering for each entry. The images that match the includes strategy take precedence over excludes matches. The image name must the includes or excludes regular expression value to be considered a match.

It is possible to modify images during image preparation, and then immediately deploy with modified images. Scenarios for modifying images include:

To modify an image during preparation, invoke an Ansible role on each image that you want to modify. The role takes a source image, makes the requested changes, and tags the result. The prepare command can push the image to the destination registry and set the heat parameters to refer to the modified image.

The Ansible role tripleo-modify-image conforms with the required role interface and provides the behaviour necessary for the modify use cases. Control the modification with the modify-specific keys in the ContainerImagePrepare parameter:

To select a use case that the tripleo-modify-image role handles, set the tasks_from variable to the required file in that role.

While developing and testing the ContainerImagePrepare entries that modify images, run the image prepare command without any additional options to confirm that the image is modified as you expect:

sudo openstack tripleo container image prepare \
  -e ~/containers-prepare-parameter.yaml

The following example ContainerImagePrepare entry updates all packages on the images using the dnf repository configuration on the undercloud host:

ContainerImagePrepare:
- push_destination: true
  ...
  modify_role: tripleo-modify-image
  modify_append_tag: "-updated"
  modify_vars:
    tasks_from: yum_update.yml
    compare_host_packages: true
    yum_repos_dir_path: /etc/yum.repos.d
  ...

You can install a directory of RPM files in your container images. This is useful for installing hotfixes, local package builds, or any package that is not available through a package repository. For example, the following ContainerImagePrepare entry installs some hotfix packages only on the nova-compute image:

ContainerImagePrepare:
- push_destination: true
  ...
  includes:
  - nova-compute
  modify_role: tripleo-modify-image
  modify_append_tag: "-hotfix"
  modify_vars:
    tasks_from: rpm_install.yml
    rpms_path: /home/stack/nova-hotfix-pkgs
  ...

For maximum flexibility, you can specify a directory containing a Dockerfile to make the required changes. When you invoke the tripleo-modify-image role, the role generates a Dockerfile.modified file that changes the FROM directive and adds extra LABEL directives. The following example runs the custom Dockerfile on the nova-compute image:

ContainerImagePrepare:
- push_destination: true
  ...
  includes:
  - nova-compute
  modify_role: tripleo-modify-image
  modify_append_tag: "-hotfix"
  modify_vars:
    tasks_from: modify_image.yml
    modify_dir_path: /home/stack/nova-custom
  ...

The following example shows the /home/stack/nova-custom/Dockerfile file. After you run any USER root directives, you must switch back to the original image default user:

FROM registry.redhat.io/rhosp-rhel8/openstack-nova-compute:latest

USER "root"

COPY customize.sh /tmp/
RUN /tmp/customize.sh

USER "nova"

Red Hat Satellite 6 offers registry synchronization capabilities that you can use to pull multiple images into a Satellite server and manage them as part of an application life cycle. The Satellite also acts as a registry for other container-enabled systems to use. For more information about managing container images, see "Managing Container Images" in the Red Hat Satellite 6 Content Management Guide.

The examples in this procedure use the hammer command line tool for Red Hat Satellite 6 and an example organization called ACME. Substitute this organization for your own Satellite 6 organization.

The following example environment file contains Satellite-specific parameters:

parameter_defaults:
  ContainerImagePrepare:
  - push_destination: false
    set:
      ceph_image: acme-production-myosp16-osp16_containers-rhceph-4
      ceph_namespace: satellite.example.com:5000
      ceph_tag: latest
      name_prefix: acme-production-myosp16-osp16_containers-
      name_suffix: ''
      namespace: satellite.example.com:5000
      neutron_driver: null
      tag: 16.0
      ...
    tag_from_label: '{version}-{release}'

You must define the containers-prepare-parameter.yaml environment file in the undercloud.conf configuration file, otherwise the undercloud uses the default values:

container_images_file = /home/stack/containers-prepare-parameter.yaml

The director installation process requires certain settings in the undercloud.conf configuration file, which director reads from the home directory of the stack user. Complete the following steps to copy default template as a foundation for your configuration.

The following list contains information about parameters for configuring the undercloud.conf file. Keep all parameters within their relevant sections to avoid errors.

[ctlplane-subnet]
cidr = 192.168.24.0/24
dhcp_start = 192.168.24.5
dhcp_end = 192.168.24.24
inspection_iprange = 192.168.24.100,192.168.24.120
gateway = 192.168.24.1
masquerade = true

You can specify as many provisioning networks as necessary to suit your environment.

Modify the values for these parameters to suit your configuration. When complete, save the file.

Complete the following steps to install director and perform some basic post-installation tasks.

The director installation is complete. You can now use the director command line tools.

The director provides commands to update the packages on the undercloud node. This allows you to perform a minor update within the current version of your OpenStack Platform environment.

This procedure demonstrates how to deploy an overcloud with minimum configuration. The result will be a basic two-node overcloud (1 Controller node, 1 Compute node).

For information on updating a containerized overcloud, see the Keeping Red Hat OpenStack Platform Updated guide.

Red Hat OpenStack Platform (RHOSP) runs services in containers on the undercloud and overcloud nodes. In certain situations, you might need to control the individual services on a host. This section contains information about some common commands you can run on a node to manage containerized services.

$ sudo podman ps

To include stopped or failed containers in the command output, add the --all option to the command:

$ sudo podman ps --all

To list container images, run the following command:

$ sudo podman images

$ sudo podman inspect keystone

To check a container status, run the systemctl status command:

$ sudo systemctl status tripleo_keystone
● tripleo_keystone.service - keystone container
   Loaded: loaded (/etc/systemd/system/tripleo_keystone.service; enabled; vendor preset: disabled)
   Active: active (running) since Fri 2019-02-15 23:53:18 UTC; 2 days ago
 Main PID: 29012 (podman)
   CGroup: /system.slice/tripleo_keystone.service
           └─29012 /usr/bin/podman start -a keystone

To stop a container, run the systemctl stop command:

$ sudo systemctl stop tripleo_keystone

To start a container, run the systemctl start command:

$ sudo systemctl start tripleo_keystone

To restart a container, run the systemctl restart command:

$ sudo systemctl restart tripleo_keystone

Because no daemon monitors the containers status, Systemd automatically restarts most containers in these situations:

For more information about Systemd services, see the systemd.service documentation.

To list all OpenStack Platform containers timers, run the systemctl list-timers command and limit the output to lines containing tripleo:

$ sudo systemctl list-timers | grep tripleo
Mon 2019-02-18 20:18:30 UTC  1s left       Mon 2019-02-18 20:17:26 UTC  1min 2s ago  tripleo_nova_metadata_healthcheck.timer            tripleo_nova_metadata_healthcheck.service
Mon 2019-02-18 20:18:33 UTC  4s left       Mon 2019-02-18 20:17:03 UTC  1min 25s ago tripleo_mistral_engine_healthcheck.timer           tripleo_mistral_engine_healthcheck.service
Mon 2019-02-18 20:18:34 UTC  5s left       Mon 2019-02-18 20:17:23 UTC  1min 5s ago  tripleo_keystone_healthcheck.timer                 tripleo_keystone_healthcheck.service
Mon 2019-02-18 20:18:35 UTC  6s left       Mon 2019-02-18 20:17:13 UTC  1min 15s ago tripleo_memcached_healthcheck.timer                tripleo_memcached_healthcheck.service
(...)

To check the status of a specific container timer, run the systemctl status command for the healthcheck service:

$ sudo systemctl status tripleo_keystone_healthcheck.service
● tripleo_keystone_healthcheck.service - keystone healthcheck
   Loaded: loaded (/etc/systemd/system/tripleo_keystone_healthcheck.service; disabled; vendor preset: disabled)
   Active: inactive (dead) since Mon 2019-02-18 20:22:46 UTC; 22s ago
  Process: 115581 ExecStart=/usr/bin/podman exec keystone /openstack/healthcheck (code=exited, status=0/SUCCESS)
 Main PID: 115581 (code=exited, status=0/SUCCESS)

Feb 18 20:22:46 undercloud.localdomain systemd[1]: Starting keystone healthcheck...
Feb 18 20:22:46 undercloud.localdomain podman[115581]: {"versions": {"values": [{"status": "stable", "updated": "2019-01-22T00:00:00Z", "..."}]}]}}
Feb 18 20:22:46 undercloud.localdomain podman[115581]: 300 192.168.24.1:35357 0.012 seconds
Feb 18 20:22:46 undercloud.localdomain systemd[1]: Started keystone healthcheck.

To stop, start, restart, and show the status of a container timer, run the relevant systemctl command against the .timer Systemd resource. For example, to check the status of the tripleo_keystone_healthcheck.timer resource, run the following command:

$ sudo systemctl status tripleo_keystone_healthcheck.timer
● tripleo_keystone_healthcheck.timer - keystone container healthcheck
   Loaded: loaded (/etc/systemd/system/tripleo_keystone_healthcheck.timer; enabled; vendor preset: disabled)
   Active: active (waiting) since Fri 2019-02-15 23:53:18 UTC; 2 days ago

If the healthcheck service is disabled but the timer for that service is present and enabled, it means that the check is currently timed out, but will be run according to timer. You can also start the check manually.

Paunch and the container-puppet.py script configure podman containers to push their outputs to the /var/log/containers/stdout directory, which creates a collection of all logs, even for the deleted containers, such as container-puppet-* containers.

The host also applies log rotation to this directory, which prevents huge files and disk space issues.

In case a container is replaced, the new container outputs to the same log file, because podman uses the container name instead of container ID.

You can also check the logs for a containerized service with the podman logs command. For example, to view the logs for the keystone container, run the following command:

$ sudo podman logs keystone

$ sudo podman exec -it keystone /bin/bash

To enter the shell for the keystone container as the root user, run the following command:

$ sudo podman exec --user 0 -it <NAME OR ID> /bin/bash

To exit the container, run the following command:

# exit

If a containerized service fails during or after overcloud deployment, use the following recommendations to determine the root cause for the failure:

$ sudo podman logs keystone

In most cases, this log provides the cause of a container’s failure.

$ sudo podman inspect keystone

This provides a JSON object containing low-level configuration data. You can pipe the output to the jq command to parse specific data. For example, to view the container mounts for the keystone container, run the following command:

$ sudo podman inspect keystone | jq .[0].Mounts

You can also use the --format option to parse data to a single line, which is useful for running commands against sets of container data. For example, to recreate the options used to run the keystone container, use the following inspect command with the --format option:

$ sudo podman inspect --format='{{range .Config.Env}} -e "{{.}}" {{end}} {{range .Mounts}} -v {{.Source}}:{{.Destination}}{{if .Mode}}:{{.Mode}}{{end}}{{end}} -ti {{.Config.Image}}' keystone

Use these options in conjunction with the podman run command to recreate the container for troubleshooting purposes:

$ OPTIONS=$( sudo podman inspect --format='{{range .Config.Env}} -e "{{.}}" {{end}} {{range .Mounts}} -v {{.Source}}:{{.Destination}}{{if .Mode}}:{{.Mode}}{{end}}{{end}} -ti {{.Config.Image}}' keystone )
$ sudo podman run --rm $OPTIONS /bin/bash

$ sudo podman exec -ti keystone <COMMAND>

Replace <COMMAND> with your desired command. For example, each container has a health check script to verify the service connection. You can run the health check script for keystone with the following command:

$ sudo podman exec -ti keystone /openstack/healthcheck

To access the container’s shell, run podman exec using /bin/bash as the command:

$ sudo podman exec -ti keystone /bin/bash

$ sudo podman export keystone -o keystone.tar

This command create the keystone.tar archive, which you can extract and explore.

The following table shows the correlation between Systemd-based services and the podman containers controlled with the Systemd services.

The following table shows Systemd-based OpenStack logs and their equivalents for containers. All container-based log locations are available on the physical host and are mounted to the container.

The following table shows Systemd-based Red Hat OpenStack Platform (RHOSP) configuration and their equivalents for containers. All container-based configuration locations are available on the physical host, are mounted to the container, and are merged, through kolla, into the configuration within each respective container.