Jump to: Complete Features | Incomplete Features | Complete Epics | Incomplete Epics | Other Complete | Other Incomplete |
Note: this page shows the Feature-Based Change Log for a release
These features were completed when this image was assembled
Feature Overview
Insights Advisor for OpenShift is integrated within OpenShift Cluster Manager. This has some limitations for adding new features and also for sharing codebase between RHEL Advisor and OCM Insights Advisor tab. Insights Advisor for OpenShift lacks certain features from the RHEL UI, the codebase is not 1:1 clone.
As a customer of Insights I will have same/very similar user experience with Insights for OpenShift and Insights for RHEL. The workflows will share the main concepts, the UI elements will be same and features introduced to Advisor will be automatically considered for both all supported platforms.
As OpenShift users I will still see integrations of Insights Advisor within OpenShift Cluster Manager that shows aggregated information for customer account and single cluster view on Advisor data. These integration will point to new Insights Advisor for OpenShift app that will be tightly integrated into OpenShift Cluster Manager.
Goals
Requirements
Benefits
Questions to answer...
Out of Scope
Background, and strategic fit
Documentation Considerations
OCP WebConsole, in the main dashboard, has an Insights Advisor widget, which has been redirecting users to OCM. Due to the Insights Advisor tab decommission in OCM, the links should point to Advisor instead.
4.10 code freeze = 28 January (marking the task as urgent)
This Feature is a general "catch all" for the time being. There are a number of existing priorities from Q1 that should be aligned with existing priorities below but if not, assign to this feature as needed.
In order to get a better overall portfolio view, we'll leverage this Feature to gather work that doesn't fall into other existing priorities on this board. As this list grows, the portfolio priority grooming team will look to split out or handle appropriately.
A list of specific needs or objectives that a Feature must deliver to satisfy the Feature. Some requirements will be flagged as MVP. If an MVP gets shifted, the feature shifts. If a non MVP requirement slips, it does not shift the feature.
requirement | Notes | isMvp |
---|---|---|
< How will the user interact with this feature? >
< Which users will use this and when will they use it? >
< Is this feature used as part of current user interface? >
< What does the person writing code, testing, documenting need to know? >
< Are there assumptions being made regarding prerequisites and dependencies?>
< Are there assumptions about hardware, software or people resources?>
< Are there specific customer environments that need to be considered (such as working with existing h/w and software)?>
< What educational or reference material (docs) is required to support this product feature? For users/admins? Other functions (security officers, etc)? >
<What does success look like?>
< Does this feature have doc impact? Possible values are: New Content, Updates to existing content, Release Note, or No Doc Impact?>
<If unsure and no Technical Writer is available, please contact Content Strategy. If yes, complete the following.>
Question | Outcome |
Console provides support UI for operators which is dynamically enabled when the operator is installed; by using feature flags against presence of CRDs. While operators have their own release cadence separately from OpenShift which makes for alignment of UI to API difficult. As new features are released for the operator, the UI becomes out of sync with APIs and customers must wait till the following OpenShift release to get any new UI.
Console extensions:
https://docs.google.com/document/d/1HW5_cl6cOX5P14PQN-1_8c60o9dMY6HbFDRftH6aTno/edit
Dynamic Plugins:
https://docs.google.com/document/d/19BAFo_8BtMZVvKsU-bE61bZpSydeYONkCMWntMU9NgE/edit
Enhancement proposal:
https://github.com/openshift/enhancements/pull/441
When this image was assembled, these features were not yet completed. Therefore, only the Jira Cards included here are part of this release
Today, all configuration for setting individual, for example, routing configuration is done via a single configuration file that only admins have access to. If an environment uses multiple tenants and each tenant, for example, has different systems that they are using to notify teams in case of an issue, then someone needs to file a request w/ an admin to add the required settings.
That can be bothersome for individual teams, since requests like that usually disappear in the backlog of an administrator. At the same time, administrators might get tons of requests that they have to look at and prioritize, which takes them away from more crucial work.
We would like to introduce a more self service approach whereas individual teams can create their own configuration for their needs w/o the administrators involvement.
Last but not least, since Monitoring is deployed as a Core service of OpenShift there are multiple restrictions that the SRE team has to apply to all OSD and ROSA clusters. One restriction is the ability for customers to use the central Alertmanager that is owned and managed by the SRE team. They can't give access to the central managed secret due to security concerns so that users can add their own routing information.
Provide a new API (based on the Operator CRD approach) as part of the Prometheus Operator that allows creating a subset of the Alertmanager configuration without touching the central Alertmanager configuration file.
Please note that we do not plan to support additional individual webhooks with this work. Customers will need to deploy their own version of the third party webhooks.
Team A wants to send all their important notifications to a specific Slack channel.
As described in https://github.com/openshift/enhancements/blob/ba3dc219eecc7799f8216e1d0234fd846522e88f/enhancements/monitoring/multi-tenant-alerting.md#distinction-between-platform-and-user-alerts, cluster admins want to distinguish platform alerts from user alerts. For this purpose, CMO should provision an external label (openshift_io_alert_source="platform") on prometheus-k8s instances.
tldr: three basic claims, the rest is explanation and one example
While bugs are an important metric, fixing bugs is different than investing in maintainability and debugability. Investing in fixing bugs will help alleviate immediate problems, but doesn't improve the ability to address future problems. You (may) get a code base with fewer bugs, but when you add a new feature, it will still be hard to debug problems and interactions. This pushes a code base towards stagnation where it gets harder and harder to add features.
One alternative is to ask teams to produce ideas for how they would improve future maintainability and debugability instead of focusing on immediate bugs. This would produce designs that make problem determination, bug resolution, and future feature additions faster over time.
I have a concrete example of one such outcome of focusing on bugs vs quality. We have resolved many bugs about communication failures with ingress by finding problems with point-to-point network communication. We have fixed the individual bugs, but have not improved the code for future debugging. In so doing, we chase many hard to diagnose problem across the stack. The alternative is to create a point-to-point network connectivity capability. this would immediately improve bug resolution and stability (detection) for kuryr, ovs, legacy sdn, network-edge, kube-apiserver, openshift-apiserver, authentication, and console. Bug fixing does not produce the same impact.
We need more investment in our future selves. Saying, "teams should reserve this" doesn't seem to be universally effective. Perhaps an approach that directly asks for designs and impacts and then follows up by placing the items directly in planning and prioritizing against PM feature requests would give teams the confidence to invest in these areas and give broad exposure to systemic problems.
Relevant links:
Plugin teams need a mechanism to extend the OCP console that is decoupled enough so they can deliver at the cadence of their projects and not be forced in to the OCP Console release timelines.
The OCP Console Dynamic Plugin Framework will enable all our plugin teams to do the following:
Requirement | Notes | isMvp? |
---|---|---|
UI to enable and disable plugins | YES | |
Dynamic Plugin Framework in place | YES | |
Testing Infra up and running | YES | |
Docs and read me for creating and testing Plugins | YES | |
CI - MUST be running successfully with test automation | This is a requirement for ALL features. | YES |
Release Technical Enablement | Provide necessary release enablement details and documents. | YES |
Documentation Considerations
Questions to be addressed:
Currently, webpack tree shakes PatternFly and only includes the components used by console in its vendor bundle. We need to expose all of the core PatternFly components for use in dynamic plugin, which means we have to disable tree shaking for PatternFly. We should expose this as a separate bundle. This will allow browsers to cache more efficiently and only need to load the PF bundle again when we upgrade PatternFly.
Open Questions
What parts of PatternFly do we consider core?
Acceptance Criteria
Requirement | Notes | isMvp? |
---|
CI - MUST be running successfully with test automation | This is a requirement for ALL features. | YES |
Release Technical Enablement | Provide necessary release enablement details and documents. | YES |
This Section:
This Section: What does the person writing code, testing, documenting need to know? What context can be provided to frame this feature.
Questions to be addressed:
As a user, I want the ability to run a pod in debug mode.
This should be the equivalent of running: oc debug pod
Acceptance Criteria for MVP
Assets
Designs (WIP): https://docs.google.com/document/d/1b2n9Ox4xDNJ6AkVsQkXc5HyG8DXJIzU8tF6IsJCiowo/edit#
OCP/Telco Definition of Done
Feature Template descriptions and documentation.
Feature Overview
Enable customers to access Google services from workloads on OpenShift clusters using Google Workload Identity (aka WIF)
https://cloud.google.com/kubernetes-engine/docs/concepts/workload-identity
Requirement | Notes | isMvp? |
---|---|---|
CI - MUST be running successfully with test automation | This is a requirement for ALL features. | YES |
Release Technical Enablement | Provide necessary release enablement details and documents. | YES |
This Section:
This Section: What does the person writing code, testing, documenting need to know? What context can be provided to frame this feature.
Questions to be addressed:
Dependencies (internal and external)
We need to ensure following things in the openshift operators
1) Make sure to operator uses v0.0.0-20210218202405-ba52d332ba99 or later version of the golang.org/x/oauth2 module
2) Mount the oidc token in the operator pod, this needs to go in the deployment. We have done it for cluster-image-registry-operator here
3) For workload identity to work, gco credentials that the operator pod uses should be of external_account type (not service_account). The external_account credentials type have path to oidc token along, url of the service account to impersonate along with other details. These type of credentials can be generated from gcp console or programmatically (supported by ccoctl). The operator pod can then consume it from a kube secret. Make appropriate code changes to the operators so that can consume these new credentials
Following repos need one or more of above changes
Enable sharing ConfigMap and Secret across namespaces
Requirement | Notes | isMvp? |
---|---|---|
Secrets and ConfigMaps can get shared across namespaces | YES |
NA
NA
Consumption of RHEL entitlements has been a challenge on OCP 4 since it moved to a cluster-based entitlement model compared to the node-based (RHEL subscription manager) entitlement mode. In order to provide a sufficiently similar experience to OCP 3, the entitlement certificates that are made available on the cluster (OCPBU-93) should be shared across namespaces in order to prevent the need for cluster admin to copy these entitlements in each namespace which leads to additional operational challenges for updating and refreshing them.
Questions to be addressed:
* What educational or reference material (docs) is required to support this product feature? For users/admins? Other functions (security officers, etc)?
* Does this feature have doc impact?
* New Content, Updates to existing content, Release Note, or No Doc Impact
* If unsure and no Technical Writer is available, please contact Content Strategy.
* What concepts do customers need to understand to be successful in [action]?
* How do we expect customers will use the feature? For what purpose(s)?
* What reference material might a customer want/need to complete [action]?
* Is there source material that can be used as reference for the Technical Writer in writing the content? If yes, please link if available.
* What is the doc impact (New Content, Updates to existing content, or Release Note)?
OCP/Telco Definition of Done
Epic Template descriptions and documentation.
<--- Cut-n-Paste the entire contents of this description into your new Epic --->
As a developer using OpenShift
I want to mount a Simple Content Access certificate into my build
So that I can access RHEL content within a Docker strategy build.
As a application developer or administrator
I want to share credentials across namespaces
So that I don't need to copy credentials to every workspace
As an OpenShift engineer
I want to know which clusters are using the Shared Resource CSI Driver
So that I can be proactive in supporting customers who are using this tech preview feature
None - metrics exported to telemetry are not formally documented.
QE can verify that the query/recording rule for cluster monitoring operator returns data if the cluster has the Shared Resource CSI driver installed and utilizes a SharedSecret or SharedConfigMap in a pod/workload.
Insights rules can potentially be created off of these exported metrics. This would allow CEE to identify which clusters are using SharedSecrets or SharedConfigMaps, especially if we are exporting mount failure metrics.
To implement, a prometheus query/recording rule needs to be added to the cluster monitoring operator. Once approved by the monitoring team, the metric data will be available on DataHub once 4.10 clusters are installed with the updated version of the monitoring operator.
As a cluster admin
I want the cluster storage operator to install the shared resources CSI driver
So that I can test the shared resources CSI driver on my cluster
Docs will need to identify how to install the shared resources CSI driver (by enabling the tech preview feature set)
Tasks:
Note that to be able to test all of this on any cloud provider, we need STOR-616 to be implemented. We can work around this by making the CSI driver installable on AWS or GCP for testing purposes.
The cluster storage operator has cluster-admin permissions. However, no other CSI driver managed by the operator includes a CRD for its API.
Upstream Kuberenetes is following other SIGs by moving it's intree cloud providers to an out of tree plugin format, Cloud Controller Manager, at some point in a future Kubernetes release. OpenShift needs to be ready to action this change
Bring together all the cloud controller managers (AWS, GCP, Azure), complete testing and prepare for final GA
A list of specific needs or objectives that a feature must deliver in order to be considered complete. Be sure to include nonfunctional requirements such as security, reliability, performance, maintainability, scalability, usability, etc. Initial completion during Refinement status.
Include use case diagrams, main success scenarios, alternative flow scenarios. Initial completion during Refinement status.
Include a list of refinement / architectural questions that may need to be answered before coding can begin. Initial completion during Refinement status.
High-level list of items that are out of scope. Initial completion during Refinement status.
Provide any additional context is needed to frame the feature. Initial completion during Refinement status.
Provide any additional customer-specific considerations that must be made when designing and delivering the Feature. Initial completion during Refinement status.
Provide information that needs to be considered and planned so that documentation will meet customer needs. Initial completion during Refinement status.
Which other projects and versions in our portfolio does this feature impact? What interoperability test scenarios should be factored by the layered products? Initial completion during Refinement status.
OCP/Telco Definition of Done
Epic Template descriptions and documentation.
<--- Cut-n-Paste the entire contents of this description into your new Epic --->
Initial work was started there: https://github.com/lobziik/cluster-cloud-controller-manager-operator/pull/1/files
Need to isolate provider specific code in respective packages and introduce interface to leverage it (regular and bootstrap manifests rendering should be there atm)
DoD:
This section includes Jira cards that are linked to an Epic, but the Epic itself is not linked to any Feature. These epics were completed when this image was assembled
An epic we can duplicate for each release to ensure we have a place to catch things we ought to be doing regularly but can tend to fall by the wayside.
Update console from Cypress 6.0.0 to 8.5.0. Changes that impact us:
https://docs.cypress.io/guides/references/migration-guide#Migrating-to-Cypress-8-0
As an adopter of the @openshift-console/dynamic-plugin-sdk I want to easily integrate into my development pipeline so that I can extend the OCP console.
Trying to pull in the dynamic-plugin-sdk into ACM is proving to be problematic. We would have to move to older dependencies. Integrating with webpack and typescript requires a very specific setup.
The dynamic-plugin-sdk has only really been used internally by OCP and is strongly tied to the setup and dependencies of OCP. For the dynamic-plugin-sdk to be externally consumable by adopters, it should be as easy to use as other webpack plugins such as HtmlWebpackPlugin or CompressionPlugin.
Update webpack to the latest 4.x and update webpack loaders. This will help prepare us to move to webpack 5.
The console has many instances of old variables, $grid-float-breakpoint and $grid-gutter-width, controlling margins/padding and responsive breakpoints throughout the Admin and Dev Console. These do not provide spacing and behaviors consistent with Patternfly components which use their own variables, $pf-global-gutter-md, $pf-global-gutter, and $pf-global-breakpoint-{size}. By replacing these, the intent it to bring the console closer to a pure Patternfly structure and behavior, requiring less overrides and customizations.
In the image-registry, we have packages origin-common and kubernetes-common. The problem is that this code doesn't get updates. We can replace them with more supported library-go.
OCP/Telco Definition of Done
Epic Template descriptions and documentation.
<--- Cut-n-Paste the entire contents of this description into your new Epic --->
As a developer using Jenkins to build my application
I want to use the base Jenkins agent image as a sidecar in my PodTemplate
So that I can use any s2i builder image in my Jenkins pipelines
QE will need to verify that the new pod templates can successfully execute a JenkinsPipeline build.
Documentation needs to be updated to explain how to use the new template.
Unclear if we need new CEE/PX materials beyond doc updates.
We currently have built-in pod templates for NodeJS and Maven, which use specialized agent images with NodeJS/Maven image.
Blog post here outlines the process: https://developers.redhat.com/blog/2020/06/04/an-easier-way-to-create-custom-jenkins-containers/
The Groovy style of declaring in-line pod templates is deprecated in favor of a YAML-style format.
Existing documentation for the Jenkin pod templates: https://docs.openshift.com/container-platform/4.9/openshift_images/using_images/images-other-jenkins.html#images-other-jenkins-config-kubernetes_images-other-jenkins
OCP/Telco Definition of Done
Epic Template descriptions and documentation.
After investigating a complex Bugzilla involving many applications making queries to prometheus-adapter, we've noticed that we were lacking insights on the requests made to prometheus-adapter. To have such information for an aggregated API, the best would be to have audit logs for prometheus-adapter. This wasn't configurable before, but with https://github.com/kubernetes-sigs/custom-metrics-apiserver/pull/92, upstream users should now be able to configure it.
Since this would greatly help in investigating prometheus-adapter Bugzilla in the future, it would be great if we allowed OpenShift users to configure the audit logs so that they could provide them to us.
Note for the assignee: as of the time of the creation of this ticket, the upstream PR hasn't been merged in custom-metrics-apiserver and thus wasn't synced in prometheus-adapter. So we will have to wait a bit before starting looking into this ticket.
DoD:
The console requires to know the network type capabilities to show/hide some Network Policy form fields.
As a result of https://issues.redhat.com/browse/NETOBSERV-27, this logic is implemented as a features document inside the console code. The console fetches the network type from the network operator and checks the supported features towards this document.
However, this limits the feature to admin users, as other logged-in users do not have permissions to fetch the network type.
This task aims to modify the current Cluster Network Operator to expose the network capabilities as an `sdn-public` Config Map, writeable only by the SDN, readable by any `system:authenticated` user.
Enhancement Proposal PR: https://github.com/openshift/enhancements/pull/875
We want to configure 'default' and 'allowed' values in validation webhook for Guest Accelerators field in GCPProviderSpec. Also revendor it to include newly added Guest Accelerators field.
This can be done after https://github.com/openshift/cluster-api-provider-gcp/pull/172 is merged.
DoD:
Description:
Openshift on RHV is composed of the following subproject the team maintains:
Each of those projects currently uses the generated oVirt API project go-ovirt.
This leads to a number of issues:
Then came go-ovirt-client, go-ovirt-client-log, go-ovirt-client-log-klog and k8sOVirtCredentialsMonitor to the rescue!
The go-ovirt-client is a wrapper around the go-ovirt which contains all the error handling/retry logic/logs/tests needed to provide a decent user experience and an easy-to-use API to the oVirt engine.
go-ovirt-client-log is a library to unify the logging logic between the projects, it is used by go-ovirt-client and should be used by all the sub-projects.
go-ovirt-client-log-klog is a companion library to go-ovirt-client-log enabling logging via the Kubernetes "klog" facility.
k8sOVirtCredentialsMonitor is a utility for monitoring the oVirt credentials secret, which will automatically update the ovirt credentials is they are changed.
We aim to move all projects which are using the go-ovirt to use go-ovirt-client, go-ovirt-client-log and k8sOVirtCredentialsMonitor instead.
Benefits for the eng:
Benefits for the customers:
Acceptance criteria:
How to test:
Description:
Acceptance:
ovirt-csi-driver uses go-ovirt-client for 95% percent of all oVirt related logic.
T-shirt size: M
Provide an easy and successful experience for front end developers to build and deploy their applications
Currently, the front end dev experience is not positive. It's much easier for them to use other platforms. Improving the front end dev experience will enable us to gain more marketshare
Although we provide the ability for 2 & 3 today, the current journey does not match with the mental model of the front end developer
Desired UX experience
As a user, I want have the option to add additional labels to a Route, as I could do in OCP3. See RFE-622
The additional labels should only be added to the route, not the service or other components. The advanced option "Labels" should not be touched and these labels are added to all components.
As an small additional we should also show always the "Target port" since it also defines the Service port and to make this more clear, the "Target port" should be shown before the "Create a route to the Application" checkbox.
The following changes should be applied to the Import flow (from Git, from Container, ...) and to the Edit page as well:
This epic is mainly focused on the 4.10 Release QE activities
1. Identify the scenarios for automation
2. Segregate the test Scenarios into smoke, Regression and other user stories
a. Update the https://docs.jboss.org/display/ODC/Automation+Status+Report
3. Align with layered operator teams for updating scripts
3. Work closely with dev team for epic automation
4. Create the automation scripts using cypress
5. Implement CI for nightly builds
6. Execute scripts on sprint basis
To the track the QE progress at one place in 4.10 Release Confluence page
There are different code spots which maps the old action items "From Git", "From Dockerfile" and "From Devfile" to the new action "Import from Git".
We should avoid mapping different strings to the new version and instead update our tests so that the feature and page object files matches the latest frontend code.
Code areas I found are marked with
// TODO (ODC-6455): Tests should use latest UI labels like "Import from Git" instead of mapping strings
This section includes Jira cards that are linked to an Epic, but the Epic itself is not linked to any Feature. These epics were not completed when this image was assembled
Please read: migrating-protractor-tests-to-cypress
Protractor test to migrate: `frontend/integration-tests/tests/oauth.scenario.ts`
Large but straight forward
47) OAuth 48) BasicAuth IDP ✔ creates a Basic Authentication IDP ✔ shows the BasicAuth IDP on the OAuth settings page 49) GitHub IDP ✔ creates a GitHub IDP ✔ shows the GitHub IDP on the OAuth settings page 50) GitLab IDP ✔ creates a GitLab IDP ✔ shows the GitLab IDP on the OAuth settings page 51) Google IDP ✔ creates a Google IDP ✔ shows the Google IDP on the OAuth settings page 52) Keystone IDP ✔ creates a Keystone IDP ✔ shows the Keystone IDP on the OAuth settings page 53) LDAP IDP ✔ creates a LDAP IDP ✔ shows the LDAP IDP on the OAuth settings page 54) OpenID IDP ✔ creates a OpenID IDP ✔ shows the OpenID IDP on the OAuth settings page
Accpetance Criteria
As a follow up to OCPCLOUD-693, we need to, once all of the API definitions are present in openshift/api, migrate the existing code bases to use the new API locations.
This will include:
Complete all the 4.9 epic features automation user stories and merge it to master branch.
4.9 epics automation completion
Tech debt should be completed
Create the pr's for 4.9 epic user stories automation
Review it
Merge it to 4.10 master branch and 4.9 master branch
As a user, I want to store my delivery pipelines in a Git repository as the source of truth and execute the pipeline on OpenShift on Git events, so that I can version and trace changes to the delivery pipelines in Git.
This section includes Jira cards that are not linked to either an Epic or a Feature. These tickets were completed when this image was assembled
console-operator codebase contains a lot of inline manifests. Instead we should put those manifests into a `/bindata` folder, from which they will be read and then updated per purpose.
Before platformStatus, the operator used to get information about AWS and GCP from the install-config config map. This code can be removed.
Description of problem:
We are seeing customer's upgrade cannot kickoff due to the availableUpdates is null in clusterversion CR
Version-Release number of the following components:
How reproducible:
sometime
Steps to Reproduce:
1.
2.
3.
Actual results:
Please include the entire output from the last TASK line through the end of output if an error is generated
Expected results:
Additional info:
The CMO e2e tests create a bunch of resources. These should be cleaned up on a successful run. However:
In a CI context this is rarely a problem, however running the tests locally can be made quite awkward, especially repeated runs on the same cluster.
We should tag all resources created by the e2e tests with a label (app.kubernetes.io/created-by: cmo-e2e-test).
This will allow easy cleanup by deleting all resources with that label and will allow for checking proper clean-up.
DoD:
All e2e resources get properly tagged.
It is straight forward to ensure that future code changes don't skip adding this tag.
Description of problem:
Previously we had a bug opened for "Reduce buildah log level for default build log level [NEEDINFO]":
https://bugzilla.redhat.com/show_bug.cgi?id=1996883
We suggested customer using secrets, however, customer confirmed that they are using secrets as per:
https://docs.openshift.com/container-platform/4.8/cicd/builds/creating-build-inputs.html#builds-input-secrets-configmaps_creating-build-inputs
But not able to use "--quiet" build argument since they are using openshift s2i config:
strategy:
type: Source
sourceStrategy:
from:
kind: ImageStreamTag
namespace: xxxxxxx
name: 's2i-xxxxxx-xxxxxxx:v1.0.0'
Actual results:
Under this setting, the secret as well as every other openshift secret are printed.
Expected results:
Sensitive information (ENV) should not appear in build logs
Additional info:
Maybe pass the --quiet option via the buildconfig fir s2i?
— Additional comment from taxu@redhat.com on 2022-07-01 03:42:46 UTC —
Hi Build team,
I see that the target release for this bug is 4.11.0
Please let us know where to add the LOG_LEVEL for s2i and/or passing "--quiet" options once the fix is deployed.
Kind regards,
Tao Xu
— Additional comment from aos-team-art-private@redhat.com on 2022-07-13 06:44:30 UTC —
Elliott changed bug status from MODIFIED to ON_QA.
This bug is expected to ship in the next 4.12 release.
— Additional comment from rdlugyhe@redhat.com on 2022-07-26 20:34:06 UTC —
Please approve the updated Doc Text.
— Additional comment from cdaley@redhat.com on 2022-07-26 20:49:05 UTC —
Approved
— Additional comment from jitsingh@redhat.com on 2022-07-27 04:48:18 UTC —
verified
— Additional comment from oarribas@redhat.com on 2022-08-15 12:11:23 UTC —
@Corey, I can see this BZ now verified for 4.12. Can this be backported to previous OCP versions?
— Additional comment from cdaley@redhat.com on 2022-08-15 13:57:02 UTC —
What version are you interested in getting it backported to?
— Additional comment from oarribas@redhat.com on 2022-08-15 15:46:10 UTC —
@Corey, I can see only 4.10 and 4.11 as "Full Support" in [1], so at least to those versions.
Acceptance criteria:
+++ This bug was initially created as a clone of Bug #2118717 +++
https://bugzilla.redhat.com/show_bug.cgi?id=2118717
Description of problem:
This BZ is a spin-off of BZ-2114945 so we can track possible issues with new TCP connections from pods failing to be created on the nodes leading to pods being unable to start or crash.
Version-Release number of selected component (if applicable):
OCP 4.10.24 with OVN-Kubernetes
How reproducible:
Periodically on the customer only so far.
— Additional comment from Andre Costa on 2022-08-10 16:30:00 UTC —
There are 3 must-gathers here that were gathered during the issues and after the restart of OVNk-masters which makes all these issues go away and pods start connections immediately.
This must-gather was taken at 11 AM today when they received a report from one of the customers:
Customer reported the issue again and this time we also got sosreport and inspect from the project.
In the pod they get errors like this (the same we saw on the call last week with them where it seems no TCP connections entries are created at all. First we though it was DNS but even with IPs directly there were issues like this):
-----------------
mx-toni-dev toni-dev-build 0/1 Error 0 18m 10.195.80.253 demchdc5vvx <none> <none>
[z0003rbj-z07@stuart ~]$ oc logs toni-dev-build
time="2022-08-10T10:59:02Z" level=info msg="Start building app with registry type openshift"
time="2022-08-10T10:59:02Z" level=info msg="Adding ssl certificate /etc/ssl/certs/ca-bundle.crt"
time="2022-08-10T10:59:02Z" level=info msg="Certificate /etc/ssl/certs/ca-bundle.crt has been added successfully"
time="2022-08-10T10:59:02Z" level=info msg="Updating docker config with registry credentials"
time="2022-08-10T10:59:02Z" level=info msg="Docker config has been updated with registry credentials"
time="2022-08-10T10:59:02Z" level=info msg="Downloading MDA from https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/eba21059-8896-4cb7-8971-4d61f6756273/71042"
time="2022-08-10T10:59:32Z" level=error msg="Failed to build mendix app, failed to create application layer failed to download MDA from https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/eba21059-8896-4cb7-8971-4d61f6756273/71042, Get \"https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/eba21059-8896-4cb7-8971-4d61f6756273/71042\": proxyconnect tcp: dial tcp: i/o timeout: Get \"https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/eba21059-8896-4cb7-8971-4d61f6756273/71042\": proxyconnect tcp: dial tcp: i/o timeout"
-----------------------
This keeps happening if they continue to run the builds which they did and created the MG and sosreport:
And like we have seen so far restarting the ovnk-master pods makes these connections work immediately again:
— Additional comment from Andre Costa on 2022-08-10 16:30:43 UTC —
— Additional comment from Tim Rozet on 2022-08-10 22:36:06 UTC —
Thanks for the must gathers. From Flavio and I examining them, there is definitely a bug here in ovn-kube. The toni-dev-build pod is deleted/recreated multiple times, and during this time it moves to different nodes. However due to a bug in OVNK, this port is updated with the new ip address and information as if it was moving to the new node, but stays on the previous logical switch. So for example, this is what happens:
1. The pod is originally assigned to node demchdc6zax. This node's cluster subnet is 10.195.79.0/24:
2022-08-09T09:22:57.078688727+00:00 stderr F I0809 09:22:57.078632 2239319 cni.go:248] [mx-toni-dev/toni-dev-build b1a4fb0be20ff717f85fd0fffab4fb303bbcb0f8b68aced4852fb7a2465d2df1] ADD finished CNI request [mx-toni-dev/toni-dev-build b1a4fb0be20ff717f85fd0fffab4fb303bbcb0f8b68aced4852fb7a2465d2df1], result "{\"interfaces\":[
{\"name\":\"b1a4fb0be20ff71\",\"mac\":\"a6:68:38:ad:66:c8\"},
{\"name\":\"eth0\",\"mac\":\"0a:58:0a:c3:4f:52\",\"sandbox\":\"/var/run/netns/e354d2d5-83cb-406f-a2d9-c5f3e786bae4\"}],\"ips\":[
{\"version\":\"4\",\"interface\":1,\"address\":\"10.195.79.82/24\",\"gateway\":\"10.195.79.1\"}],\"dns\":{}}", err <nil
2. Over time this pod is completed, deleted, recreated many times. Until eventually it lands on demchdc5vvx the next day:
2022-08-10T08:43:58.428759111Z I0810 08:43:58.428719 1837017 cni.go:248] [mx-toni-dev/toni-dev-build 5d4f195cbda5269e5451593987be9d69ea828ee549bc447a5bbe50db847c182a] ADD finished CNI request [mx-toni-dev/toni-dev-build 5d4f195cbda5269e5451593987be9d69ea828ee549bc447a5bbe50db847c182a], result "{\"interfaces\":[
,
{\"name\":\"eth0\",\"mac\":\"0a:58:0a:c3:50:1e\",\"sandbox\":\"/var/run/netns/bae8f77a-b368-4b3a-86dc-df925330fa26\"}],\"ips\":[
{\"version\":\"4\",\"interface\":1,\"address\":\"10.195.80.30/24\",\"gateway\":\"10.195.80.1\"}],\"dns\":{}}", err <nil>
3. Although it lands on a new node, in OVNK we update the old port (somehow the old port is not being removed) that is attached to the old switch:
[root@fedora ~]# ovn-nbctl list logical_switch_port c345cc07-8a89-4e70-beff-d8d9f4dac46a
_uuid : c345cc07-8a89-4e70-beff-d8d9f4dac46a
addresses : ["0a:58:0a:c3:50:1e 10.195.80.30"]
dhcpv4_options : []
dhcpv6_options : []
dynamic_addresses : []
enabled : []
external_ids :
ha_chassis_group : []
name : mx-toni-dev_toni-dev-build
options :
parent_name : []
port_security : ["0a:58:0a:c3:50:1e 10.195.80.30"]
tag : []
tag_request : []
type : ""
up : false
[root@fedora ~]# ovn-nbctl lsp-list demchdc6zax | grep c345cc07-8a89-4e70-beff-d8d9f4dac46a
c345cc07-8a89-4e70-beff-d8d9f4dac46a (mx-toni-dev_toni-dev-build)
[root@fedora ~]# ovn-nbctl lsp-list demchdc5vvx | grep c345cc07-8a89-4e70-beff-d8d9f4dac46a
[root@fedora ~]#
This will cause the pod not to be able to send any traffic as its IP is in the wrong subnet for this switch.
4. Additionally the default node SNAT for this pod is in the right place:
[root@fedora ~]# ovn-nbctl lr-nat-list GR_demchdc6zax | grep 10.195.80.30
[root@fedora ~]# ovn-nbctl lr-nat-list GR_demchdc5vvx | grep 10.195.80.30
snat 139.25.144.25 10.195.80.30
5. But there is no egress IP reroute or SNAT entry for this pod:
Egress IP:
status:
items:
[root@fedora ~]# ovn-nbctl lr-nat-list GR_demchdc5z6x | grep 10.195.80.30
[root@fedora ~]# ovn-nbctl lr-nat-list GR_demchdc5z6x | grep 139.25.144.72
snat 139.25.144.72 10.195.77.156
snat 139.25.144.72 10.195.80.40
snat 139.25.144.72 10.195.76.65
snat 139.25.144.72 10.195.76.184
snat 139.25.144.72 10.195.80.42
snat 139.25.144.72 10.195.80.156
6. We see in the ovnkube-master logs that ovnk attempts to delete this pod, but it fails because we try to delete a logical switch port that is still bound to the wrong logical switch:
2022-08-10T09:07:32.033303027Z I0810 09:07:32.033270 1 client.go:781] "msg"="transacting operations" "database"="OVN_Northbound" "operations"="[{Op:mutate Table:Address_Set Row:map[] Rows:[] Columns:[] Mutations:[{Column:addresses Mutator:delete Value:{GoSet:[10.195.80.30]}}] Timeout:<nil> Where:[where column _uuid ==
] Until: Durable:<nil> Comment:<nil> Lock:<nil> UUIDName:} {Op:mutate Table:Logical_Switch Row:map[] Rows:[] Columns:[] Mutations:[{Column:ports Mutator:delete Value:{GoSet:[
{GoUUID:c345cc07-8a89-4e70-beff-d8d9f4dac46a}]}}] Timeout:<nil> Where:[where column _uuid == {f5073eaa-3f72-4ec2-94c3-3744d412864a}] Until: Durable:<nil> Comment:<nil> Lock:<nil> UUIDName:} {Op:delete Table:Logical_Switch_Port Row:map[] Rows:[] Columns:[] Mutations:[] Timeout:<nil> Where:[where column _uuid == {c345cc07-8a89-4e70-beff-d8d9f4dac46a}] Until: Durable:<nil> Comment:<nil> Lock:<nil> UUIDName:}]"Rows:[]} {Count:1 Error: Details: UUID:
{GoUUID:} Rows:[]} {Count:1 Error: Details: UUID:{GoUUID}Rows:[]} {Count:0 Error:referential integrity violation Details:cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s) UUID:
{GoUUID:} Rows:[]}] and errors []: referential integrity violation: cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s)Rows:[]} {Count:1 Error: Details: UUID:
{GoUUID:} Rows:[]} {Count:1 Error: Details: UUID:{GoUUID}Rows:[]} {Count:0 Error:referential integrity violation Details:cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s) UUID:
{GoUUID:} Rows:[]}] and errors []: referential integrity violation: cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s)Rows:[]} {Count:1 Error: Details: UUID:
{GoUUID:} Rows:[]} {Count:1 Error: Details: UUID:{GoUUID}Rows:[]} {Count:0 Error:referential integrity violation Details:cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s) UUID:
{GoUUID:} Rows:[]}] and errors []: referential integrity violation: cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s)Rows:[]} {Count:1 Error: Details: UUID:
{GoUUID:} Rows:[]} {Count:1 Error: Details: UUID:{GoUUID}Rows:[]} {Count:0 Error:referential integrity violation Details:cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s) UUID:
{GoUUID:} Rows:[]}] and errors []: referential integrity violation: cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s)Rows:[]} {Count:1 Error: Details: UUID:
{GoUUID:} Rows:[]} {Count:1 Error: Details: UUID:{GoUUID}Rows:[]} {Count:0 Error:referential integrity violation Details:cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s) UUID:
{GoUUID:}Rows:[]}] and errors []: referential integrity violation: cannot delete Logical_Switch_Port row d300fdb7-d337-4c64-8e31-7ff02889d9fb because of 1 remaining reference(s)
2022-08-11T08:26:24.264204783Z time="2022-08-11T08:26:24Z" level=error msg="Failed to build mendix app, failed to create application layer failed to download MDA from https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/ef924516-7121-4372-a306-b0055c445766/71268, Get \"https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/ef924516-7121-4372-a306-b0055c445766/71268\": proxyconnect tcp: dial tcp: i/o timeout: Get \"https://privatecloud.mendixcloud.com/rest/mdarepository/v1/download/ef924516-7121-4372-a306-b0055c445766/71268\": proxyconnect tcp: dial tcp: i/o timeout"
— Additional comment from Tim Rozet on 2022-08-11 21:19:12 UTC —
We were able to reproduce the issue locally. There are two potential paths that can cause a stale logical switch port to be re-used on the wrong node:
Scenario 1: pod is created, deleted, and recreated on another node extremely quickly. This plays out like this:
Events:
1. pod toni is created on node A
2. pod toni is deleted on node A
3. pod toni is recreated on node B
What happens in ovnk:
1. pod toni is created on node B
2. pod toni is deleted on node A (fails)
This happens because we grab the latest version of the pod to add in event 1, which by the time we grab it is actually the value from event 3. Then after processing event 1, we move to event 2 and when we delete the pod, we use what was given to us in the event, which is now incorrect to remove the pod from node A, because it actually is on node B. The fix is to ignore the value in the pod spec on deletion, and use what we store internally in our internal port cache. If there is no entry in the cache, we search OVN NBDB to find the right switch (this is an expensive operation so we want to avoid it when possible).
Flavio is working on a fix for this.
Scenario 2: pod is created, runs to completion, is deleted very quickly, and then is recreated on another node:
Events:
1. pod toni is created on node A
2. pod toni runs to completion
3. pod toni immediately is deleted
4. pod toni is recreated on node B
What happens in ovnk:
1. pod toni is created on node A
2. completion causes an update event, ovnk does not delete the pod (bug)
3. deletion event is processed, but since it is a completed pod, we ignore it (since we should have already delted it in step 2)
4. pod toni is recreated on node B, ovnk sees there is already an switch port for toni on the wrong switch, and just updates that with the new information
This happens when a pod goes to completed, but is deleted very quickly. In step 2 during update event we try to grab the latest version of the pod, but it doesn't exist anymore since it was deleted. In this case we skip the update, instead of tearing down the pod. The delete code in step 3 assumes that if the pod is completed we must have already handled it in update, so the stale port stays around until an add later re-uses it. Patryk already has a fix for this:
https://github.com/ovn-org/ovn-kubernetes/pull/3071
There may still be more egress IP issues to investigate (tracked in other bugs) after fixing this, and we will look into those after fixing these fundamental pod issues.
— Additional comment from Tim Rozet on 2022-08-11 22:19:30 UTC —
Andre, re: comment 5. This looks like the same issue. If you have a must gather we can confirm. I think the sosreport from the worker node is not enough. The referential integrity violation occurs because we attempt to do these operations:
1. remove the logical switch port from the logical switch
2. delete the logical switch port
In this case:
1. we remove the logical switch port from the logical switch, but the switch port actually exists on a different switch (node) - this is a noop
2. we try to delete the logical switch port- NBDB complains this is an violation and refuses to delete it, because there is another switch that holds a reference to this object
— Additional comment from Anurag saxena on 2022-08-12 21:16:56 UTC —
@rbrattai@redhat.com Can you help verifying this and backports? feel free to re-assign to someone else if needed while I am on PTO. Thanks!
+++ This bug was initially created as a clone of
Bug #2070318
+++
Description of problem:
In OCP VRRP deployment (using OCP cluster networking), we have an additional data interface which is configured along with the regular management interface in each control node. In some deployments, the kubernetes address 172.30.0.1:443 is nat’ed to the data management interface instead of the mgmt interface (10.40.1.4:6443 vs 10.30.1.4:6443 as we configure the boostrap node) even though the default route is set to 10.30.1.0 network. Because of that, all requests to 172.30.0.1:443 were failed. After 10-15 minutes, OCP magically fixes it and nat’ing correctly to 10.30.1.4:6443.
Version-Release number of selected component (if applicable):
How reproducible:
Steps to Reproduce:
1.Provision OCP cluster using cluster networking for DNS & Load Balancer instead of external DNS & Load Balancer. Provision the host with 1 management interface and an additional interface for data network. Along with OCP manifest, add manifest to create a pod which will trigger communication with kube-apiserver.
2.Start cluster installation.
3.Check on the custom pod log in the cluster when the first 2 master nodes were installing to see GET operation to kube-apiserver timed out. Check nft table and chase the ip chains to see the that the data IP address was nat'ed to kubernetes service IP address instead of the management IP. This is not happening all the time, we have seen 50:50 chance.
Actual results:
After 10-15 minutes OCP will correct that by itself.
Expected results:
Wrong natting should not happen.
Additional info:
ClusterID: 24bbde0b-79b3-4ae6-afc5-cb694fa48895
ClusterVersion: Stable at "4.8.29"
ClusterOperators:
clusteroperator/authentication is not available (OAuthServerRouteEndpointAccessibleControllerAvailable: Get "
https://oauth-openshift.apps.ocp-binhle-wqepch.contrail.juniper.net/healthz
": context deadline exceeded (Client.Timeout exceeded while awaiting headers)) because OAuthServerRouteEndpointAccessibleControllerDegraded: Get "
https://oauth-openshift.apps.ocp-binhle-wqepch.contrail.juniper.net/healthz
": context deadline exceeded (Client.Timeout exceeded while awaiting headers)
clusteroperator/baremetal is degraded because metal3 deployment inaccessible
clusteroperator/console is not available (RouteHealthAvailable: failed to GET route (
https://console-openshift-console.apps.ocp-binhle-wqepch.contrail.juniper.net/health
): Get "
https://console-openshift-console.apps.ocp-binhle-wqepch.contrail.juniper.net/health
": context deadline exceeded (Client.Timeout exceeded while awaiting headers)) because RouteHealthDegraded: failed to GET route (
https://console-openshift-console.apps.ocp-binhle-wqepch.contrail.juniper.net/health
): Get "
https://console-openshift-console.apps.ocp-binhle-wqepch.contrail.juniper.net/health
": context deadline exceeded (Client.Timeout exceeded while awaiting headers)
clusteroperator/dns is progressing: DNS "default" reports Progressing=True: "Have 4 available DNS pods, want 5."
clusteroperator/ingress is degraded because The "default" ingress controller reports Degraded=True: DegradedConditions: One or more other status conditions indicate a degraded state: CanaryChecksSucceeding=False (CanaryChecksRepetitiveFailures: Canary route checks for the default ingress controller are failing)
clusteroperator/insights is degraded because Unable to report: unable to build request to connect to Insights server: Post "
https://cloud.redhat.com/api/ingress/v1/upload
": dial tcp: lookup cloud.redhat.com on 172.30.0.10:53: read udp 10.128.0.26:53697->172.30.0.10:53: i/o timeout
clusteroperator/network is progressing: DaemonSet "openshift-network-diagnostics/network-check-target" is not available (awaiting 1 nodes)
— Additional comment from
bnemec@redhat.com
on 2022-03-30 20:00:25 UTC —
This is not managed by runtimecfg, but in order to route the bug correctly I need to know which CNI plugin you're using - OpenShiftSDN or OVNKubernetes. Thanks.
— Additional comment from
lpbinh@gmail.com
on 2022-03-31 08:09:11 UTC —
Hi Ben,
We were deploying Contrail CNI with OCP. However, this issue happens at very early deployment time, right after the bootstrap node is started
and there's no SDN/CNI there yet.
— Additional comment from
bnemec@redhat.com
on 2022-03-31 15:26:23 UTC —
Okay, I'm just going to send this to the SDN team then. They'll be able to provide more useful input than I can.
— Additional comment from
trozet@redhat.com
on 2022-04-04 15:22:21 UTC —
Can you please provide the iptables rules causing the DNAT as well as the routes on the host? Might be easiest to get a sosreport during initial bring up during that 10-15 min when the problem occurs.
— Additional comment from
lpbinh@gmail.com
on 2022-04-05 16:45:13 UTC —
All nodes have two interfaces:
eth0: 10.30.1.0/24
eth1: 10.40.1.0/24
machineNetwork is 10.30.1.0/24
default route points to 10.30.1.1
The kubeapi service ip is 172.30.0.1:443
all Kubernetes services are supposed to be reachable via machineNetwork (10.30.1.0/24)
To make the kubeapi service ip reachable in hostnetwork, something (openshift installer?) creates a set of nat rules which translates the service ip to the real ip of the nodes which have kubeapi active.
Initially kubeapi is only active on the bootstrap node so there should be a nat rule like
172.30.0.1:443 -> 10.30.1.10:6443 (assuming that 10.30.1.10 is the bootstrap nodes' ip address in the machine network)
However, what we see is
172.30.0.1:443 -> 10.40.1.10:6443 (which is the bootstrap nodes' eth1 ip address)
The rule is configured on the controller nodes and lead to asymmetrical routing as the controller sends a packet FROM machineNetwork (10.30.1.x) to 172.30.0.1 which is then translated and forwarded to 10.40.1.10 which then tries to reply back on the 10.40.1.0 network which fails as the request came from 10.30.1.0 network.
So, we want to understand why openshift installer picks the 10.40.1.x ip address rather than the 10.30.1.x ip for the nat rule. What's the mechanism for getting the ip in case the system has multiple interfaces with ips configured.
Note: after a while (10-20 minutes) the bootstrap process resets itself and then it picks the correct ip address from the machineNetwork and things start to work.
— Additional comment from
smerrow@redhat.com
on 2022-04-13 13:55:04 UTC —
Note from Juniper regarding requested SOS report:
In reference to
https://bugzilla.redhat.com/show_bug.cgi?id=2070318
that @Binh Le has been working on. The mustgather was too big to upload for this Bugzilla. Can you access this link?
https://junipernetworks-my.sharepoint.com/:u:/g/personal/sleigon_juniper_net/ETOrHMqao1tLm10Gmq9rzikB09H5OUwQWZRAuiOvx1nZpQ
— Additional comment from
smerrow@redhat.com
on 2022-04-21 12:24:33 UTC —
Can we please get an update on this BZ?
Do let us know if there is any other information needed.
— Additional comment from
trozet@redhat.com
on 2022-04-21 14:06:00 UTC —
Can you please provide another link to the sosreport? Looks like the link is dead.
— Additional comment from
smerrow@redhat.com
on 2022-04-21 19:01:39 UTC —
See mustgather here:
https://drive.google.com/file/d/16y9IfLAs7rtO-SMphbYBPgSbR4od5hcQ
— Additional comment from
trozet@redhat.com
on 2022-04-21 20:57:24 UTC —
Looking at the must-gather I think your iptables rules are most likely coming from the fact that kube-proxy is installed:
[trozet@fedora must-gather.local.288458111102725709]$ omg get pods -n openshift-kube-proxy
NAME READY STATUS RESTARTS AGE
openshift-kube-proxy-kmm2p 2/2 Running 0 19h
openshift-kube-proxy-m2dz7 2/2 Running 0 16h
openshift-kube-proxy-s9p9g 2/2 Running 1 19h
openshift-kube-proxy-skrcv 2/2 Running 0 19h
openshift-kube-proxy-z4kjj 2/2 Running 0 19h
I'm not sure why this is installed. Is it intentional? I don't see the configuration in CNO to enable kube-proxy. Anyway the node IP detection is done via:
https://github.com/kubernetes/kubernetes/blob/f173d01c011c3574dea73a6fa3e20b0ab94531bb/cmd/kube-proxy/app/server.go#L844
Which just looks at the IP of the node. During bare metal install a VIP is chosen and used with keepalived for kubelet to have kapi access. I don't think there is any NAT rule for services until CNO comes up. So I suspect what really is happening is your node IP is changing during install, and kube-proxy is getting deployed (either intentionally or unintentionally) and that is causing the behavior you see. The node IP is chosen via the node ip configuration service:
https://github.com/openshift/machine-config-operator/blob/da6494c26c643826f44fbc005f26e0dfd10513ae/templates/common/_base/units/nodeip-configuration.service.yaml
This service will determine the node ip via which interfaces have a default route and which one has the lowest metric. With your 2 interfaces, do they both have default routes? If so, are they using dhcp and perhaps its random which route gets installed with a lower metric?
— Additional comment from
trozet@redhat.com
on 2022-04-21 21:13:15 UTC —
Correction: looks like standalone kube-proxy is installed by default when the provider is not SDN, OVN, or kuryr so this looks like the correct default behavior for kube-proxy to be deployed.
— Additional comment from
lpbinh@gmail.com
on 2022-04-25 04:05:14 UTC —
Hi Tim,
You are right, kube-proxy is deployed by default and we don't change that behavior.
There is only 1 default route configured for the management interface (10.30.1.x) , we used to have a default route for the data/vrrp interface (10.40.1.x) with higher metric before. As said, we don't have the default route for the second interface any more but still encounter the issue pretty often.
— Additional comment from
trozet@redhat.com
on 2022-04-25 14:24:05 UTC —
Binh, can you please provide a sosreport for one of the nodes that shows this behavior? Then we can try to figure out what is going on with the interfaces and the node ip service. Thanks.
— Additional comment from
trozet@redhat.com
on 2022-04-25 16:12:04 UTC —
Actually Ben reminded me that the invalid endpoint is actually the boostrap node itself:
172.30.0.1:443 -> 10.30.1.10:6443 (assuming that 10.30.1.10 is the bootstrap nodes' ip address in the machine network)
vs
172.30.0.1:443 -> 10.40.1.10:6443 (which is the bootstrap nodes' eth1 ip address)
So maybe a sosreport off that node is necessary? I'm not as familiar with the bare metal install process, moving back to Ben.
— Additional comment from
lpbinh@gmail.com
on 2022-04-26 08:33:45 UTC —
Created attachment 1875023 [details]sosreport
— Additional comment from
lpbinh@gmail.com
on 2022-04-26 08:34:59 UTC —
Created attachment 1875024 [details]sosreport-part2
Hi Tim,
We observe this issue when deploying clusters using OpenStack instances as our infrastructure is based on OpenStack.
I followed the steps here to collect the sosreport:
https://docs.openshift.com/container-platform/4.8/support/gathering-cluster-data.html
Got the sosreport which is 22MB which exceeds the size permitted (19MB), so I split it to 2 files (xaa and xab), if you can't join them then we will need to put the collected sosreport on a share drive like we did with the must-gather data.
Here are some notes about the cluster:
First two control nodes are below, ocp-binhle-8dvald-ctrl-3 is the bootstrap node.
[core@ocp-binhle-8dvald-ctrl-2 ~]$ oc get node
NAME STATUS ROLES AGE VERSION
ocp-binhle-8dvald-ctrl-1 Ready master 14m v1.21.8+ed4d8fd
ocp-binhle-8dvald-ctrl-2 Ready master 22m v1.21.8+ed4d8fd
We see the behavior that wrong nat'ing was done at the beginning, then corrected later:
sh-4.4# nft list table ip nat | grep 172.30.0.1
meta l4proto tcp ip daddr 172.30.0.1 tcp dport 443 counter packets 3 bytes 180 jump KUBE-SVC-NPX46M4PTMTKRN6Y
sh-4.4# nft list chain ip nat KUBE-SVC-NPX46M4PTMTKRN6Y
table ip nat {
chain KUBE-SVC-NPX46M4PTMTKRN6Y
}
sh-4.4# nft list chain ip nat KUBE-SEP-VZ2X7DROOLWBXBJ4
table ip nat {
chain KUBE-SEP-VZ2X7DROOLWBXBJ4
}
sh-4.4#
sh-4.4#
<....after a while....>
sh-4.4# nft list chain ip nat KUBE-SVC-NPX46M4PTMTKRN6Y
table ip nat {
chain KUBE-SVC-NPX46M4PTMTKRN6Y
}
sh-4.4# nft list chain ip nat KUBE-SEP-X33IBTDFOZRR6ONM
table ip nat {
chain KUBE-SEP-X33IBTDFOZRR6ONM
}
sh-4.4#
— Additional comment from
lpbinh@gmail.com
on 2022-05-12 17:46:51 UTC —
@
trozet@redhat.com
May we have an update on the fix, or the plan for the fix? Thank you.
— Additional comment from
lpbinh@gmail.com
on 2022-05-18 21:27:45 UTC —
Created support Case 03223143.
— Additional comment from
vkochuku@redhat.com
on 2022-05-31 16:09:47 UTC —
Hello Team,
Any update on this?
Thanks,
Vinu K
— Additional comment from
smerrow@redhat.com
on 2022-05-31 17:28:54 UTC —
This issue is causing delays in Juniper's CI/CD pipeline and makes for a less than ideal user experience for deployments.
I'm getting a lot of pressure from the partner on this for an update and progress. I've had them open a case [1] to help progress.
Please let us know if there is any other data needed by Juniper or if there is anything I can do to help move this forward.
[1]
https://access.redhat.com/support/cases/#/case/03223143
— Additional comment from
vpickard@redhat.com
on 2022-06-02 22:14:23 UTC —
@
bnemec@redhat.com
Tim mentioned in
https://bugzilla.redhat.com/show_bug.cgi?id=2070318#c14
that this issue appears to be at BM install time. Is this something you can help with, or do we need help from the BM install team?
— Additional comment from
bnemec@redhat.com
on 2022-06-03 18:15:17 UTC —
Sorry, I missed that this came back to me.
(In reply to Binh Le from
comment #16
)> We observe this issue when deploying clusters using OpenStack instances as
> our infrastructure is based on OpenStack.This does not match the configuration in the must-gathers provided so far, which are baremetal. Are we talking about the same environments?
I'm currently discussing this with some other internal teams because I'm unfamiliar with this type of bootstrap setup. I need to understand what the intended behavior is before we decide on a path forward.
— Additional comment from
rurena@redhat.com
on 2022-06-06 14:36:54 UTC —
(In reply to Ben Nemec from
comment #22
)> Sorry, I missed that this came back to me.
>
> (In reply to Binh Le from comment #16)
> > We observe this issue when deploying clusters using OpenStack instances as
> > our infrastructure is based on OpenStack.
>
> This does not match the configuration in the must-gathers provided so far,
> which are baremetal. Are we talking about the same environments?
>
> I'm currently discussing this with some other internal teams because I'm
> unfamiliar with this type of bootstrap setup. I need to understand what the
> intended behavior is before we decide on a path forward.I spoke to the CU they tell me that all work should be on baremetal. They were probably just testing on OSP and pointing out that they saw the same behavior.
— Additional comment from
bnemec@redhat.com
on 2022-06-06 16:19:37 UTC —
Okay, I see now that this is an assisted installer deployment. Can we get the cluster ID assigned by AI so we can take a look at the logs on our side? Thanks.
— Additional comment from
lpbinh@gmail.com
on 2022-06-06 16:38:56 UTC —
Here is the cluster ID, copied from the bug description:
ClusterID: 24bbde0b-79b3-4ae6-afc5-cb694fa48895
In regard to your earlier question about OpenStack & baremetal (2022-06-03 18:15:17 UTC):
We had an issue with platform validation in OpenStack earlier. Host validation was failing with the error message “Platform network settings: Platform OpenStack Compute is allowed only for Single Node OpenShift or user-managed networking.”
It's found out that there is no platform type "OpenStack" available in [
https://github.com/openshift/assisted-service/blob/master/models/platform_type.go#L29
] so we set "baremetal" as the platform type on our computes. That's the reason why you are seeing baremetal as the platform type.
Thank you
— Additional comment from
ercohen@redhat.com
on 2022-06-08 08:00:18 UTC —
Hey, first you are currect, When you set 10.30.1.0/24 as the machine network, the bootstrap process should use the IP on that subnet in the bootstrap node.
I'm trying to understand how exactly this cluster was installed.
You are using on-prem deployment of assisted-installer (podman/ACM)?
You are trying to form a cluster from OpenStack Vms?
You set the platform to Baremetal where?
Did you set user-managed-netwroking?
Some more info, when using OpenStack platform you should install the cluster with user-managed-netwroking.
And that's what the failing validation is for.
— Additional comment from
bnemec@redhat.com
on 2022-06-08 14:56:53 UTC —
Moving to the assisted-installer component for further investigation.
— Additional comment from
lpbinh@gmail.com
on 2022-06-09 07:37:54 UTC —
@Eran Cohen:
Please see my response inline.
You are using on-prem deployment of assisted-installer (podman/ACM)?
--> Yes, we are using on-prem deployment of assisted-installer.
You are trying to form a cluster from OpenStack Vms?
--> Yes.
You set the platform to Baremetal where?
--> It was set in the Cluster object, Platform field when we model the cluster.
Did you set user-managed-netwroking?
--> Yes, we set it to false for VRRP.
— Additional comment from
itsoiref@redhat.com
on 2022-06-09 08:17:23 UTC —
@
lpbinh@gmail.com
can you please share assisted logs that you can download when cluster is failed or installed?
Will help us to see the full picture
— Additional comment from
ercohen@redhat.com
on 2022-06-09 08:23:18 UTC —
OK, as noted before when using OpenStack platform you should install the cluster with user-managed-netwroking (set to true).
Can you explain how you workaround this failing validation? “Platform network settings: Platform OpenStack Compute is allowed only for Single Node OpenShift or user-managed networking.”
What does this mean exactly? 'we set "baremetal" as the platform type on our computes'
To be honest I'm surprised that the installation was completed successfully.
@
oamizur@redhat.com
I thought installing on OpenStack VMs with baremetal platform (user-managed-networking=false) will always fail?
— Additional comment from
lpbinh@gmail.com
on 2022-06-10 16:04:56 UTC —
@
itsoiref@redhat.com
: I will reproduce and collect the logs. Is that supposed to be included in the provided must-gather?
@
ercohen@redhat.com
:
— Additional comment from
itsoiref@redhat.com
on 2022-06-13 13:08:17 UTC —
@
lpbinh@gmail.com
you will have download_logs link in UI. Those logs are not part of must-gather
— Additional comment from
lpbinh@gmail.com
on 2022-06-14 18:52:02 UTC —
Created attachment 1889993 [details]cluster log per need info request - Cluster ID caa475b0-df04-4c52-8ad9-abfed1509506
Attached is the cluster log per need info request.
Cluster ID: caa475b0-df04-4c52-8ad9-abfed1509506
In this reproduction, the issue is not resolved by OpenShift itself, wrong NAT still remained and cluster deployment failed eventually
sh-4.4# nft list table ip nat | grep 172.30.0.1
meta l4proto tcp ip daddr 172.30.0.1 tcp dport 443 counter packets 2 bytes 120 jump KUBE-SVC-NPX46M4PTMTKRN6Y
sh-4.4# nft list chain ip nat KUBE-SVC-NPX46M4PTMTKRN6Y
table ip nat {
chain KUBE-SVC-NPX46M4PTMTKRN6Y
}
sh-4.4# nft list chain ip nat KUBE-SEP-VZ2X7DROOLWBXBJ4; date
table ip nat {
chain KUBE-SEP-VZ2X7DROOLWBXBJ4
}
Tue Jun 14 17:40:19 UTC 2022
sh-4.4# nft list chain ip nat KUBE-SEP-VZ2X7DROOLWBXBJ4; date
table ip nat {
chain KUBE-SEP-VZ2X7DROOLWBXBJ4
}
Tue Jun 14 17:59:19 UTC 2022
sh-4.4# nft list chain ip nat KUBE-SEP-VZ2X7DROOLWBXBJ4; date
table ip nat {
chain KUBE-SEP-VZ2X7DROOLWBXBJ4
}
Tue Jun 14 18:17:38 UTC 2022
sh-4.4#
sh-4.4#
sh-4.4# nft list chain ip nat KUBE-SEP-VZ2X7DROOLWBXBJ4; date
table ip nat {
chain KUBE-SEP-VZ2X7DROOLWBXBJ4
}
Tue Jun 14 18:49:28 UTC 2022
sh-4.4#
— Additional comment from
itsoiref@redhat.com
on 2022-06-15 15:59:22 UTC —
@
lpbinh@gmail.com
just for the protocol, we don't support baremetal ocp on openstack that's why validation is failing
— Additional comment from
lpbinh@gmail.com
on 2022-06-15 17:47:39 UTC —
@
itsoiref@redhat.com
as explained it's just a workaround on our side to make OCP work in our lab, and from my understanding on OCP perspective it will see that deployment is on baremetal only, not related to OpenStack (please correct me if I am wrong).
We have been doing thousands of OCP cluster deployments in our automation so far, if it's why validation is failing then it should be failing every time. However it only occurs occasionally when nodes have 2 interfaces, using OCP internal DNS and Load balancer, and sometime resolved by itself and sometime not.
— Additional comment from
itsoiref@redhat.com
on 2022-06-19 17:00:01 UTC —
For now i can assume that this endpoint is causing the issue:
{
"apiVersion": "v1",
"kind": "Endpoints",
"metadata": {
"creationTimestamp": "2022-06-14T17:31:10Z",
"labels":
,
"name": "kubernetes",
"namespace": "default",
"resourceVersion": "265",
"uid": "d8f558be-bb68-44ac-b7c2-85ca7a0fdab3"
},
"subsets": [
{
"addresses": [
],
"ports": [
{
"name": "https",
"port": 6443,
"protocol": "TCP"
}
]
}
]
},
— Additional comment from
itsoiref@redhat.com
on 2022-06-21 17:03:51 UTC —
The issue is that kube-api service advertise wrong ip but it does it cause kubelet chooses the one arbitrary and we currently have no mechanism to set kubelet ip, especially in bootstrap flow.
— Additional comment from
lpbinh@gmail.com
on 2022-06-22 16:07:29 UTC —
@
itsoiref@redhat.com
how do you perform OCP deployment in setups that have multiple interfaces if letting kubelet chooses an interface arbitrary instead of configuring a specific IP address for it to listen on? With what you describe above chance of deployment failure in system with multiple interfaces would be high.
— Additional comment from
dhellard@redhat.com
on 2022-06-24 16:32:26 UTC —
I set the Customer Escalation flag = Yes, per ACE EN-52253.
The impact is noted by the RH Account team: "Juniper is pressing and this impacts the Unica Next Project at Telefónica Spain. Unica Next is a critical project for Red Hat. We go live the 1st of July and this issue could impact the go live dates. We need clear information about the status and its possible resolution.
— Additional comment from
itsoiref@redhat.com
on 2022-06-26 07:28:44 UTC —
I have sent an image with possible fix to Juniper and waiting for their feedback, once they will confirm it works for them we will proceed with the PRs.
— Additional comment from
pratshar@redhat.com
on 2022-06-30 13:26:26 UTC —
=== In Red Hat Customer Portal Case 03223143 ===
— Comment by Prateeksha Sharma on 6/30/2022 6:56 PM —
//EMT note//
Update from our consultant Manuel Martinez Briceno -
====
on 28th June, 2022 the last feedback from Juniper Project Manager and our Partner Manager was that they are testing the fix. They didn't give an Estimate Time to finish, but we will be tracking this closely and let us know of any news.
====
Thanks & Regards,
Prateeksha Sharma
Escalation Manager | RHCSA
Global Support Services, Red Hat
Description of problem:
OCP Upgrade failing
Version-Release number of the following components:
oc version
Client Version: 4.8.0-202108312109.p0.git.0d10c3f.assembly.stream-0d10c3f
Server Version: 4.10.13
Kubernetes Version: v1.23.5+b463d71
How reproducible: Always
Steps to Reproduce:
1. Create the following SCC (that has `with readOnlyRootFilesystem: true`):
~~~
cat << EOF | oc create -f -
allowHostDirVolumePlugin: true
allowHostIPC: false
allowHostNetwork: false
allowHostPID: false
allowHostPorts: false
allowPrivilegeEscalation: true
allowPrivilegedContainer: true
allowedCapabilities: []
apiVersion: security.openshift.io/v1
defaultAddCapabilities: []
fsGroup:
type: MustRunAs
groups: []
kind: SecurityContextConstraints
metadata:
annotations:
meta.helm.sh/release-name: azure-arc
meta.helm.sh/release-namespace: default
labels:
app.kubernetes.io/managed-by: Helm
name: kube-aad-proxy-scc
priority: null
readOnlyRootFilesystem: true
requiredDropCapabilities: []
runAsUser:
type: RunAsAny
seLinuxContext:
type: MustRunAs
supplementalGroups:
type: RunAsAny
users:
2. oc adm upgrade --to=4.10.20
Actual results:
SCC kube-aad-proxy-scc, which has readOnlyRootFilesystem is injected inside the pod version-4.10.20-smvt9-6vqwc, causing it to fail.
~~~
Expected results:
Pod version-4.10.20-smvt9-6vqwc should run fine
Additional info:
I don't know why, but SCC kube-aad-proxy-scc is injected inside pod version-4.10.20-smvt9-6vqwc:
~~~
apiVersion: v1
kind: Pod
metadata:
annotations:
k8s.v1.cni.cncf.io/network-status: |-
[{
"name": "openshift-sdn",
"interface": "eth0",
"ips": [
"10.129.0.70"
],
"default": true,
"dns": {}
}]
k8s.v1.cni.cncf.io/networks-status: |-
[{
"name": "openshift-sdn",
"interface": "eth0",
"ips": [
"10.129.0.70"
],
"default": true,
"dns": {}
}]
openshift.io/scc: kube-aad-proxy-scc ### HERE
creationTimestamp: "2022-07-25T16:47:39Z"
generateName: version-4.10.20-5xqtv-
labels:
controller-uid: ba707bbe-1825-4f80-89ce-f6bf2301a812
job-name: version-4.10.20-5xqtv
name: version-4.10.20-5xqtv-9gcwk
namespace: openshift-cluster-version
ownerReferences:
*USER STORY:*
As a customer or OpenShift engineer, I want to see the user agent for anything calling from OpenShift -> vSphere to eliminate troubleshooting guesswork.
*DESCRIPTION:*
A question in #forum-vmware was raised where we identified that the user-agent may not be configured for all OpenShift components calling to vSphere API.
https://coreos.slack.com/archives/CH06KMDRV/p1627368902058800
*Required:*
Audit of OpenShift components calling to vSphere API to make sure user agent strings are set appropriately.
*Nice to have:*
How can this be prevented in the future? How can we minimize maintenance costs added by new PRs/bugs reported from this spike?
*ACCEPTANCE CRITERIA:*
New PRs or bug reports for each effected component.
As per [1], the jsonnet code for managing thanos-ruler resources should reuse the upstream kube-thanos project.
Goal
We have several use cases where dynamic plugins need to proxy to another service on the cluster. One example is the Helm plugin. We would like to move the backend code for Helm to a separate service on the cluster, and the Helm plugin could proxy to that service for its requests. This is required to make Helm a dynamic plugin. Similarly if we want to have ACM contribute any views through dynamic plugins, we will need a way for ACM to proxy to its services (e.g., for Search).
It's possible for plugins to make requests to services exposed through routes today, but that has several problems:
Plugins need a way to declare in-cluster services that they need to connect to. The console backend will need to set up proxies to those services on console load. This also requires that the console operator be updated to pass the configuration to the console backend.
This work will apply only to single clusters.
Open Questions
Acceptance Criteria
cc Ali Mobrem [~christianmvogt]
Currently, Telemeter is not equipped with configurable request limit for receive endpoint (for full context see: https://github.com/openshift/cluster-monitoring-operator/pull/1416). It is using the default limit defined in the code base, however it seems this limit might not be suitable for our usage.
As a part of this ticket, it should be:
1) Understood what is the appropriate limit for request size for our use cases
2) Make the limit configurable in Telemeter via a flag
3) Deploy the changes, initially to the staging environment, to enable our team to test it.
Add a Makefile rule in CMO to execute all the different rule that are used for verification and validation. Currenctly, some of them might not be at the right place, for example `check-assets` which is part of `generate` despite not being responsible of any generation. https://github.com/openshift/cluster-monitoring-operator/pull/1151/files#r629371735
DoD:
The static authorizer feature has landed in upstream kube-rbac-proxy. Lets use it by configuring a static authorizer for all requests that hit a /metrics endpoint.
DoD:
As mentioned in [1], the cluster monitoring operator doesn't define the relatedObjects field in the ClusterOperator manifest which is initially deployed by CVO [2].
If the CMO pod fails to start, the must-gather might miss information from the monitoring namespace. Note that once CMO runs, it will update the initial ClusterOperator object with the proper information [3].
[1] http://mailman-int.corp.redhat.com/archives/aos-devel/2021-May/msg00139.html
[2] https://github.com/openshift/cluster-monitoring-operator/blob/master/manifests/0000_50_cluster-monitoring-operator_06-clusteroperator.yaml
[3] https://github.com/openshift/cluster-monitoring-operator/blob/a6bc9824035ceb8dbfe7c53cf0c138bfb2ec5643/pkg/client/status_reporter.go#L49-L63
This section includes Jira cards that are not linked to either an Epic or a Feature. These tickets were not completed when this image was assembled
The current integration of prometheus-adapter in OpenShift uses the platform Prometheus as a backend to get metrics. The problem with this design is that we are getting metrics from 2 different Prometheus instances which don't have replicated data, so two queries sent at the same time to prometheus-adapter might yield different results since the underlying promQL queries executed by prometheus-adapter might be on different Prometheus servers. The consequence is that we end up having inconsistent data across multiple autoscaling requests.
This can be easily tested by running:
$ while true ; do date; oc adm top pod -n openshift-monitoring prometheus-k8s-0 ; echo; sleep 1 ;done Mon Jul 26 03:55:07 EDT 2021 NAME CPU(cores) MEMORY(bytes) prometheus-k8s-0 208m 4879Mi Mon Jul 26 03:55:08 EDT 2021 NAME CPU(cores) MEMORY(bytes) prometheus-k8s-0 246m 4877Mi Mon Jul 26 03:55:09 EDT 2021 NAME CPU(cores) MEMORY(bytes) prometheus-k8s-0 208m 4879Mi Mon Jul 26 03:55:10 EDT 2021 NAME CPU(cores) MEMORY(bytes) prometheus-k8s-0 246m 4877Mi
This isn't a bug in itself since it was designed that way, but we could do better by using thanos-querier as a backend instead of the platform Prometheus because it will duplicate the metrics from both instances and serve one consistent result based on the data that it will get from the Prometheuses.
DoD: