dci-openshift-app-agent
enables Cloud-Native Applications and Operators in OpenShift using Red Hat Distributed CI service.
This agent is expected to be installed in a RHEL8 server (from now on referred as jumphost) with access to the API of an already deployed OCP cluster.
Before starting make sure the next list of items are covered in the jumphost server.
- Be running the latest stable RHEL release (8.4 or higher) and registered via RHSM
- Ansible 2.9 (See section Newer Ansible Versions for newer Ansible versions)
- Access to the Internet, it could be through a proxy (See section Proxy Considerations)
- Access to the following repositories:
- epel
- dci-release
- baseos-rpms
- appstream-rpms
- Podman 3.0 (See section Old Podman versions for older Podman versions)
- Kubernetes Python module
- An OpensShift cluster already deployed or a process to deploy it before running the
dci-openshift-app-agent
.
In an already registered server with RHEL you can fulfil the repositories and Ansible 2.9 requirements with the following commands:
# subscription-manager repos --enable=rhel-8-for-x86_64-baseos-rpms
# subscription-manager repos --enable=rhel-8-for-x86_64-appstream-rpms
# subscription-manager repos --enable ansible-2.9-for-rhel-8-x86_64-rpms
# dnf -y install https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
# dnf -y install https://packages.distributed-ci.io/dci-release.el8.noarch.rpm
The dci-openshift-app-agent
is packaged and available as a RPM file located in this repository. It can be installed in the jumphost server with the following command:
# dnf -y install dci-openshift-app-agent
To configure your DCI job pipelines, you need to install dci-pipeline
. Instructions at dci-pipeline documentation.
Once dci-openshift-app-agent
package is installed, the files and resources you can find in this repository will be placed in the following locations:
/etc/dci-openshift-app-agent
contains these folders and files:dcirc.sh.dist
file,hooks
folder,hosts.yml
file andsettings.yml
file./etc/sysconfig
folder contains the content ofsysconfig
folder, which isdci-openshift-app-agent
file./usr/share/dci-openshift-app-agent/
gathers the following folders and files:ansible.cfg
file,dci-openshift-app-agent.yml
file,group_vars
folder,plays
folder andutilities
folder./var/lib/dci-openshift-app-agent
folder holds thesamples
folder./usr/bin
folder holds scripts such asdci-openshift-app-agent-ctl
.
Note: scripts provided in this agent are deprecated. You should use
dci-pipeline
instead.
Also, have in mind that:
dci-openshift-app-agent
user (with sudo permissions) and group are created- Files under
systemd
folder in this repo will be used to create the corresponding system service fordci-openshift-app-agent
.
The DCI dashboard gives you a view into what jobs you have run in your distributed agent. In order to gain access to it you have to:
-
Go to https://www.distributed-ci.io/ and click login. You will be redirected to sso.redhat.com so you'll use your RH account credentials.
-
If you are not part of any team you can contact an admin to get yourself added.
-
You will have to create a Remote CI for using it later, go on the left navigation bar on the
Remotecis
option and click. on "Create a new remoteci" -
Fill out the description and which team it belongs to, then click
Create
. -
You should see your newly created remoteci in the list, you can get the credentials in YAML format by click the button in the Authentication column. This should be saved under
~/.config/dci-pipeline/dci_credentials.yml
.This file allows to provide some variables to the DCI OpenShift App Agent. The table below shows the available variables and their default values.
Name | Default | Description |
---|---|---|
dci_config_dir | Path of the desired hooks directory | |
dci_comment | Comment to associate with the job | |
dci_url | URL to associate with the job | |
dci_disconnected | false | Set it to true if it is a disconnected environment. This variable is not defined in the default values, but all the checks that make use of this variable has a default value (i.e. false) defined. |
dci_local_registry | "" | Registry to fetch containers that may be used. Mandatory for disconnected environments. |
provision_cache_store | "/opt/cache" | Directory aimed to share artifacts between dci-openshift-agent and dci-openshift-app-agent. |
partner_creds | "" | Authfile with registries' creds. This variable must have a value if we are on a disconnected environment. In that case, it must include the creds for the local registry used, and optionally have private registry creds for partners. If it is a connected environment, this variable is optional and, if it has a value, the file would contain just private registry creds for partners. In this link, there are examples about how the files should be formatted. This variable has to be defined when running preflight on Red Hat Best Practices Test Suite for Kubernetes. |
dci_workarounds | [] | List of workarounds to be considered in the execution. Each element of the list must be a String with the following format: bz<id> or gh-org-repo-<id> |
dci_openshift_app_image | Image that can be to be used on the agent workloads. It needs to be defined in the partner hooks. It needs to be mirrored to a local registry in disconnected environments. | |
dci_openshift_app_ns | "myns" | Default namespace to deploy workloads in the running cluster. |
dci_must_gather_images | ["registry.redhat.io/openshift4/ose-must-gather"] | List of the must-gather images to use when retrieving logs. |
provisioner_name | Provisioner address (name or IP) to be accessed for retrieving logs with must-gather images. If not defined, logs will not be retrieved. | |
provisioner_user | Provisioner username, used to access to the provisioner for retrieving logs with must-gather images. If not defined, logs will not be retrieved. | |
dci_ga_components_for_certification | ["ocp"] | list of components that needs to be ga to submit a certification record |
do_certsuite | false | Enable/Disable the Red Hat Best Practices Test Suite for Kubernetes (https://github.com/redhat-best-practices-for-k8s/certsuite) |
do_chart_verifier | false | Enable/Disable the Chart Verifier |
do_must_gather | true | Enable/Disable the generation of must_gather |
do_preflight_tests | false | Trigger to activate the preflight tests |
do_container_parallel_test | false | If true, Run Preflight test in parallel for container that need to recertify or without submission to the portal |
tests_to_verify | undefined | List of expected test results. When defined, it triggers the validation of actual test results against the expectations. Please check verify-tests readme to get more details and an example of the configuration. |
See Operator Certification (preflight) for details to enable the Operator Certifications tests suite | ||
See k8s_best_practices_certsuite role for details to enable the Red Hat Best Practices Test Suite for Kubernetes | ||
See chart-verifier role for details to enable the chart-verifier tests | ||
See resources-to-components role for details to enable the creation of DCI components based on Kubernetes resources deployed in the cluster, making use of rtc_resources_to_components variable. |
Here is an example of a pipeline job definition for dci-openshift-app-agent
:
- name: ocp-workload
stage: workload
prev_stages: [ocp-upgrade, ocp]
ansible_playbook: /usr/share/dci-openshift-app-agent/dci-openshift-app-agent.yml
ansible_cfg: ~/my-lab-config/pipelines/ansible.cfg
ansible_inventory: ~/my-lab-config/inventories/@QUEUE/@RESOURCE-workload.yml
dci_credentials: ~/.config/dci-pipeline/dci_credentials.yml
ansible_extravars:
dci_cache_dir: ~/dci-cache-dir
dci_config_dir: ~/my-lab-config/ocp-workload
dci_gits_to_components:
- ~/my-lab-config
dci_local_log_dir: ~/upload-errors
dci_tags: []
use_previous_topic: true
inputs:
kubeconfig: kubeconfig_path
The DCI App Agent has support to execute multiple test suites to validate containers, virtual functions, Helm charts, and operators. The suites are in the form of Ansible roles executed during the Red Hat testing phases. The suites help the partners on getting prepared for the Red Hat Certifications or actually run the certification process on the the workloads deployed via DCI.
This test suite will run the command line interface for validating if OpenShift Operator Bundles and images meet minimum requirements for Red Hat Operator Certification.
For specific details about the features and variables for this test suite see: Preflight role documentation.
The test results for each operator can be submitted to the Red Hat connect Site. Please see Preflight Role documentation about how to enable this feature.
The k8s_best_practices_certsuite role allows to run the certsuite tests in order to meet minimum requirements for Red Hat OpenShift Certification.
For specific details about the features and variables for this test suite see: k8s_best_practices_certsuite role documentation.
Also, the tnf_test_example sample can be followed as a good example of how to deploy a workload to be tested by the Red Hat Best Practices Test Suite for Kubernetes.
Helm Chart Verifier is a test tool that validates Helm charts based on Red Hat recommendations.
The chart_verifier role is able to deploy charts on an OCP cluster and run the helm chart verifier tests. Please see the role documentation for more details about how to run tests using via the app agent or using a pipeline.
The test suites are executed in the following order, in the case all them are enabled.
- Helm chart erifier
- Red Hat Best Practices Test Suite for Kubernetes
- Preflight container only
- Preflight operators
- Operator SDK
Test suites may requires some delay between each execution. The following variables allow setting a pause between the execution of each test. The time must be set in minutes.
chart_verified_wait: 0
certsuite_wait: 0
If the execution order or pre-defined workflow does not suits the partner needs, it it recommended to use agent hooks that will allow to use the roles available for the App agent and define a custom execution workflow. See tnf_test_example sample for a reference about the hooks structure.
The dci-openshift-app-agent
is an Ansible playbook that enables Cloud-Native Applications and Operators in OpenShift using Red Hat Distributed CI service. The main entrypoint is the file dci-openshift-app-agent.yml
. It is composed of several steps executed sequentially.
In case of an issue, the agent will terminate its execution by launching (optionally) the teardown and failure playbooks.
There are two fail statuses:
failure
- Whenever there's an issue with either the installation of the application or during testing.error
- Whenever there's an error anywhere else during the playbook.
All the tasks prefixed with test_ will get exported in Junit using the Ansible Junit callback and submitted automatically to the DCI control server.
Files located in /etc/dci-openshift-app-agent/hooks/
need to be filled by the user.
They will NOT be replaced when the dci-openshift-app-agent
RPM is updated.
The hooks that can be defined are the following:
├── hooks
│  ├── pre-run.yml
│  ├── install.yml
│  ├── tests.yml
│  ├── post-run.yml
│  └── teardown.yml
This hook is used for preparation steps required in the jumphost
or inside the cluster. For example, it could be used to install some packages required in the jumphost
or to prepare workload pre-requisites.
This hook is not required and can be omitted.
Tags:
pre-run
This is the main hook that must take care of install and/or configure the application in the cluster.
This hook is required and will fail if not available.
Tags:
install
running
There is one special stage defined between the install and tests stages, called post-install, which is defined to apply configurations that must be done between both stages. This is not an official hook that can be used by users, but it is an internal play for the agent.
The tests hook is used to test the application in the install hook, this hook verifies and/or validates the install step in the cluster.
This hook is not required and can be omitted, but is highly recommended to define a way to validate/verify the installed application.
Tags:
testing
running
This hook is used for tasks required after the application has been installed and/or validated. For example, to upload logs to a central location, or to report the state of the application.
This hook is not required and can be omitted.
You can put your logs in {{ job_logs.path }}
and they will be uploaded in plays/post-run.yml
NOTE: Test result files must follow the Junit format, must be stored within the
{{ job_logs.path }}
directory and the file name must follow the pattern*.xml
.
Tags:
post-run
While this hook is not required it is highly recommended to be included. Its main purpose is to remove or destroy anything that was created with the install and/or the test hooks.
This hook is controlled with two variables:
dci_teardown_on_success
(default: true)dci_teardown_on_failure
(default: false)
It's included either when there's a failure, error or at the end of all the steps.
Some examples of hooks are provided in the $HOME directory of the dci-openshift-app-agent
user (/var/lib/dci-openshift-app-agent/samples/). You can use those to initialize the agent tests.
To use these samples, you need to include the variable dci_config_dir
with the path of the sample to use in you pipeline job definition.
NOTE: Please check the README.md for more information of how to use the examples.
-
To create a namespace and a webserver pod, to validate it is running, and to delete it, the pipeline job definition will look like this:
workload-pipeline.yml
ansible_extravars: dci_openshift_app_ns: testns dci_config_dir: /var/lib/dci-openshift-app-agent/samples/control_plane_example
You can store secrets in an encrypted manner in your pipeline definition files and YAML inventories by using dci-vault
to generate your encrypted secrets. Details in the python-dciclient documentation.
A DCI job produces a set of relevant configuration files, logs, reports, and test results that are collected during the last execution stages. The following table depicts the most relevant.
File | Section | Description |
---|---|---|
all-nodes.yaml | Files | The output oc get get nodes command |
<pod_name>.log | Files | The log entries for a given pod |
<namespace>_events.log | Files | The OCP events collected for an specific namespace |
<namespace>_status.log | Files | The list of pods that were deployed in an specific namespace |
*.log | Files | Log files generated during the job execution and stored in the dci_log directory |
*.trace | Files | Tracing files generated during the job execution and stored in the dci_log directory |
clusternetwork.yaml | Files | File describing the network configuration of the cluster |
clusteroperator.txt | Files | Report of the status of the cluster operators |
dci-openshift-agent- | Files | dci-openshift-agent tests report as JUnit format |
clusterversion.txt | Files | Report of the OCP version applied to the cluster |
events.txt | Files | Output of the oc get events -A command |
nodes.txt | Files | Output of the oc get nodes -o yaml command |
must_gather.tar.gz | Files | Debugging information about your cluster, it can be used for support cases or for troubleshooting using the O Must Gather tool |
pods.txt | Files | Output of the oc get pods -A command |
certsuite_config.yml | Files | The config file passed to certsuite (if enabled) |
certsuite-stdout.log and certsuite.log | Files | The output of certsuite execution, for troubleshooting purposes (if enabled) |
certsuite-tests_junit.xml | Files | JUnit file gathering certsuite results (if enabled) |
claim.json | Files | Claim file generated by the certsuite (if enabled) |
<preflight_*> | Files | Multiple logs, reports, and test results generated by the execution of the preflight test suite (if enabled) |
<chart-name>_*report.yaml | Files | Helm chart verifier report for the chart under test |
helm-submission-report.txt | Files | Report of helm charts submitted for certification (if enabled) |
*junit | Tests | Processed JUnit files generated by the Job or partner tests |
apirequestcounts_removed_api.csv | Files | This file lists the Cluster APIs used by a workload that have been marked for deprecation and removed in upcoming OCP versions |
apirequestcounts_ocp_compatibility.xml | Files | The compatibility of the workload with OCP versions |
version.txt | Files | Report of the OCP client and server version using during the deployment |
diff-jobs.txt | Files | A post-run stage to report that compares the current job versus the previous job of the same type regarding the components used |
operators.json | Files | A JSON file with details about the operators installed in the cluster |
If you want to test the Red Hat Best Practices Test Suite for Kubernetes in a libvirt environment, remember to tag the OCP nodes to fit in the NodeSelector property defined in partner's pod (NodeSelectors: role=partner
):
for master in $(oc get node -l node-role.kubernetes.io/master -o=custom-columns=name:.metadata.name --no-headers); do \
oc label node ${master} role=partner \
done
Red Hat Enterprise Linux 8 defaults to Ansible 2.9 installed from Ansible or base repositories, it is highly recommended to use this version because using Ansible >= 2.10 requires newer versions of other python modules that can affect your entire server. You can install Ansible 2.10 or Ansible core (2.11) via other methods, but because starting at 2.10 the use of collections has introduced some changes, you need to install a few collections to make it work
After installing the agent, login as dci-openshift-app-agent user and install the following collections
# su - dci-openshift-app-agent
$ ansible-galaxy collection install community.kubernetes
$ ansible-galaxy collection install community.general
Also the use of newer Ansible versions requires a recent version of the kubernetes python module ( >= 12.0.0), as today only available through pip3
You can upgrade the current version for the dci-openshift-app-agent user only or install a specific version like this:
$ python3 -m pip install -U kubernetes --user
# or
$ python3 -m pip install kubernetes==12.0.1 --user
In this section from the Ansible documentation web page, you can find the instructions to be followed to upgrade Ansible to a newer version.
Apart from that, it is recommended to do the following:
- Remove all Ansible packages installed via dnf
- Remove dci-* packages if already installed
- Reboot the jumphost (i.e. the server where dci-openshift-app-agent is executed)
- Reinstall Ansible from pip/pip3 as recommended in the documentation
- Reinstall dci packages
And check that everything works.
You might encounter an error when running the dci-openshift-app-agent for first time with the message Topic: XYZ Resource not found
the reasons could be the following:
- Incorrect spelling of the Topic. Login to the DCI web dashboard and go to the
Topics
section in the left menu. There you could find the correct names and Topics available to use. - If spelling is correct, then this might be a permissions issue. Ask the partner team administrator to verify the permissions on the group your account belongs.
Conditions in which the issue appeared:
- dci-openshift-app-agent installed.
- Execution of dci-openshift-app-agent directly using dci-openshift-app-agent-ctl, with the dci-openshift-app-agent user.
- Attempt to run a container, using podman, in localhost (e.g. certsuite container for running the Red Hat Best Practices Test Suite for Kubernetes).
Under these conditions, the error presented is the following (there may be other different errors, but all related to the same issue - lack of IDs available):
Error processing tar file(exit status 1): there might not be enough IDs available in the namespace (requested 0:5 for /usr/bin/write): lchown /usr/bin/write: invalid argument
Error: unable to pull quay.io/redhat-best-practices-for-k8s/certsuite:unstable: unable to pull image: Error committing the finished image: error adding layer with blob "sha256:0a3cf4c29951bdca5c283957249a78290fb441c4ef2ce74f51815056e4be7e7f": Error processing tar file(exit status 1): there might not be enough IDs available in the namespace (requested 0:5 for /usr/bin/write): lchown /usr/bin/write: invalid argument
The problem seems to be related to the subordinate user and group mapping applied for the dci-openshift-app-agent user, a feature that is needed to run rootless containers in podman, or to isolate containers with a user namespace.
This issue has already been fixed by removing -r option when creating the dci-openshift-app-agent user in dci-openshift-app-agent.spec file. But, in case you installed dci-openshift-app-agent prior to this patch (you can check it by looking at /etc/subuid and /etc/subgid files, confirming that no entries related to dci-openshift-app-agent user are present there), you have to follow these steps:
-
Check the content of /etc/subuid and /etc/subgid files. If you have already installed dci-openshift-agent on your system, you should have an entry like this in both files:
dci-openshift-agent:100000:65536
-
Copy that entry and paste it in both files, but setting the first value to dci-openshift-app-agent. If dci-openshift-agent is not installed in your server, then copy directly the line above and change the first value to dci-openshift-app-agent. Both /etc/subuid and /etc/subgid files should include a line like this now:
dci-openshift-app-agent:100000:65536
-
To conclude, execute the podman system migration command, it will take care of killing the podman "pause" process and restarting the running containers with the new subuid/subgid mapping.
podman system migrate
If the containers do not restart automatically, then you can try to restart them manually.
Make sure of changing the ownership of certain resources (e.g. the ones under /var/lib/dci-openshift-app-agent directory):
sudo chown \ -R \ dci-openshift-app-agent:dci-openshift-app-agent \ /var/lib/dci-openshift-app-agent/
Then, you can retry to deploy the containers and it should work.
References:
- https://access.redhat.com/solutions/4381691
- https://docs.docker.com/engine/security/userns-remap/
- https://www.redhat.com/sysadmin/debug-rootless-podman-mounted-volumes
It is recommended to use, at least, Podman 3.0 when running dci-openshift-app-agent, to avoid issues when pulling or building container images.
If sudo dnf update podman
does not work, you may need to follow some of these steps:
-
Try to clean the cache just in case, and then list the available packages from the upstream repository to confirm if version 3 is listed. For example, for Podman 1.6.4, it is indeed provided by
rhel-8-for-x86_64-appstream-rpms
repo:$ sudo dnf clean all $ sudo dnf --disableexcludes all --disablerepo all --enablerepo rhel-8-for-x86_64-appstream-rpms list --available --showduplicates podman # If version 3 is listed, then run again: $ sudo dnf update podman
-
Try to find issues related to the Podman version you are using and apply the workarounds proposed. For example, for Podman 1.6.4, there was a bug related to the usage of rootless containers, in which it seems that, by removing .config and .local directories and starting from scratch, it fixed the issue.
-
Confirm that the container tools version that corresponds to Podman 3 is enabled or not, and enable it if disabled (as long as it does not impact in other workloads apart from dci-openshift-app-agent running in the server). It can be checked with the following command:
$ sudo dnf module list container-tools # If container-tools rhel8 is not enabled, type the following: $ sudo dnf module reset container-tools $ sudo dnf module enable container-tools:rhel8 # And then upgrade podman
When running a job, DCI app agent interacts with the API through a module called dci-auth
, and it uses HMAC auth v4 to verify data is correct and authentic. First a signature is computed on the client side using the UTC date of the server, then another signature is computed again on DCI server. To avoid a replay attack, the signature is only valid for a few minutes. If you are running a job and see MAC errors, for example: Hmac2Mechanism failed: signature is expired
, the server clock where the DCI agent is executed may be desynchronized.
A method to confirm this is the issue, it's to execute the command dcictl topic-list
after sourcing the remote-ci credentials to list the topics, you might see None
as result:
$ source /etc/dci-openshift-app-agent/dcirc.sh
$ dcictl topic-list
+------+------+-------+-----------------+----------------+------------+
| id | name | state | component_types | export_control | product_id |
+------+------+-------+-----------------+----------------+------------+
| None | None | None | None | None | None |
+------+------+-------+-----------------+----------------+------------+
The solution is to sync the system clock to an NTP server, the time zone is not important, use timedatectl
command to validate if the clock is synchronized with an NTP service and enable accordingly if not.
$ timedatectl
Local time: Mon 2022-12-05 13:13:00 CST
Universal time: Mon 2022-12-05 19:13:00 UTC
RTC time: Mon 2022-12-05 19:13:00
Time zone: America/Chicago (CST, -0600)
System clock synchronized: yes
NTP service: active
RTC in local TZ: no
NOTE: If you move the system clock around 20 minutes forward you can replicate this issue.
If you use a proxy to go to the Internet, export the following variables in the dci-openshift-app-agent user session where you run the agent, if you use the systemd service then it would be a good idea to store these variables in the ~/.bashrc file of the dci-openshift-app-agent user
Replace PROXY-IP:PORT with your respective settings and 10.X.Y.Z/24 with the cluster subnet of the OCP cluster, and example.com with the base domain of your cluster.
export http_proxy=http://PROXY-IP:PORT
export https_proxy=http://PROXY-IP:PORT
export no_proxy=10.X.Y.Z/24,.example.com
NOTE: Also consider setting the proxy settings in the /etc/rhsm/rhsm.conf file if you use the Red Hat CDN to pull packages, otherwise the agent might fail to install dependencies required during the execution of the CNF test suite.
If you want to test a code change from Gerrit, you need to have dci-check-change
installed on your system from the dci-openshift-agent
package.
Then for example, if you want to test the change from https://softwarefactory-project.io/r/c/dci-openshift-app-agent/+/22647, issue the following command:
dci-check-change 22647 /var/lib/dci-openshift-agent/clusterconfigs/kubeconfig
You can omit the kubeconfig file as a second argument if you want dci-check-change
to re-install OCP using dci-openshift-agent-ctl
before testing the change.
All the other arguments will be passed to dci-openshift-app-agent-ctl
.
Apache License, Version 2.0 (see LICENSE file)
Email: Distributed-CI Team [email protected] IRC: #distributed-ci on Freenode