Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[EKSCTL create cluster command] Authorise self-managed nodes via aws-auth configmap when EKS access entries are disabled #7698

Merged

Conversation

TiberiuGC
Copy link
Collaborator

@TiberiuGC TiberiuGC commented Apr 2, 2024

Description

This PR fixes an inconsistent behaviour between eksctl create cluster and eksctl create nodegroup commands. Previously, when executing the former, self-managed nodegroup authorisation was always achieved via EKS access entries, regardless how accessConfig.AuthenticationMode was set. For situations where the aforementioned field was set to CONFIG_MAP (i.e. EKS access entries are disabled), self-managed nodegroup creation fails with error below when trying to create a corresponding access entry:

Resource handler returned message: "The cluster's authentication mode must be set to one of [API, API_AND_CONFIG_MAP] to perform this operation.

Therefore we now align eksctl create cluster behaviour with eksctl create nodegroup, i.e. authorising self-managed nodes via aws-auth configmap when EKS access entries are disabled.

Fixes #7695

Checklist

  • Added tests that cover your change (if possible)
  • Added/modified documentation as required (such as the README.md, or the userdocs directory)
  • Manually tested
  • Made sure the title of the PR is a good description that can go into the release notes
  • (Core team) Added labels for change area (e.g. area/nodegroup) and kind (e.g. kind/improvement)

BONUS POINTS checklist: complete for good vibes and maybe prizes?! 🤯

  • Backfilled missing tests for code in same general area 🎉
  • Refactored something and made the world a better place 🌟

@TiberiuGC TiberiuGC force-pushed the bugfix/self-managed-without-access-entries branch from 7a33223 to 79a6b41 Compare April 2, 2024 12:00
@TiberiuGC TiberiuGC changed the title Disable access entry creation for self-managed nodes on clusters with CONFIG_MAP only Disable access entry creation for self-managed nodes on clusters with authenticationMode:CONFIG_MAP Apr 2, 2024
Copy link
Member

@yuxiang-zhang yuxiang-zhang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

some sort of testing would be nice (even manual)

a-hilaly
a-hilaly previously approved these changes Apr 2, 2024
Copy link
Member

@a-hilaly a-hilaly left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM 👍

@@ -44,7 +46,8 @@ func (c *StackCollection) NewTasksToCreateCluster(ctx context.Context, nodeGroup
Parallel: true,
IsSubTask: true,
}
if unmanagedNodeGroupTasks := c.NewUnmanagedNodeGroupTask(ctx, nodeGroups, false, false, false, vpcImporter); unmanagedNodeGroupTasks.Len() > 0 {
disableAccessEntryCreation := accessConfig.AuthenticationMode == ekstypes.AuthenticationModeConfigMap
if unmanagedNodeGroupTasks := c.NewUnmanagedNodeGroupTask(ctx, nodeGroups, false, false, disableAccessEntryCreation, vpcImporter); unmanagedNodeGroupTasks.Len() > 0 {
Copy link
Member

@a-hilaly a-hilaly Apr 2, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are we testing NewUnmanagedNodeGroupTask function with true/false disableAccessEntryCreation ?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

NewUnmanagedNodeGroupTask is just passing it to other structs, then to other functions, as per the overall tasks architecture. disableAccessEntryCreation is being unit tested when it is eventually used by the CFN stack builder.

Integration tests shall cover the different behaviours though.

@TiberiuGC
Copy link
Collaborator Author

some sort of testing would be nice (even manual)

You're right! This needs a bit more work, I've found another related issue setting up the auth configmap which I'll tackle in this PR.

@TiberiuGC TiberiuGC marked this pull request as draft April 2, 2024 19:13
@TiberiuGC TiberiuGC force-pushed the bugfix/self-managed-without-access-entries branch from a726656 to d64a798 Compare April 3, 2024 09:49
@TiberiuGC TiberiuGC changed the title Disable access entry creation for self-managed nodes on clusters with authenticationMode:CONFIG_MAP Allow self-managed nodes to join the cluster during eksctl create cluster command when authenticationMode:CONFIG_MAP Apr 3, 2024
@TiberiuGC TiberiuGC changed the title Allow self-managed nodes to join the cluster during eksctl create cluster command when authenticationMode:CONFIG_MAP [EKSCTL create cluster command] Authorise self-managed nodes via aws-auth configmap when EKS access entries are disabled Apr 3, 2024
@TiberiuGC TiberiuGC force-pushed the bugfix/self-managed-without-access-entries branch from d64a798 to 1895911 Compare April 3, 2024 13:42
@TiberiuGC TiberiuGC marked this pull request as ready for review April 4, 2024 13:44
Copy link
Member

@a-hilaly a-hilaly left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍

@TiberiuGC TiberiuGC force-pushed the bugfix/self-managed-without-access-entries branch from 1895911 to 9fc4aec Compare April 4, 2024 16:58
@TiberiuGC TiberiuGC enabled auto-merge (squash) April 4, 2024 17:06
@TiberiuGC TiberiuGC merged commit 2addd3a into eksctl-io:main Apr 4, 2024
9 checks passed
hspencer77 pushed a commit to brave-intl/eksctl that referenced this pull request Apr 30, 2024
…-auth configmap` when EKS access entries are disabled (eksctl-io#7698)

* Disable access entry creation for self-managed nodes on clusters with CONFIG_MAP only

* fix logic for updating aws-auth configmap
hspencer77 added a commit to brave-intl/eksctl that referenced this pull request Apr 30, 2024
* Safely access/mutate fargate coredns pod annotations

Prior to this patch, the `pkg/fargate/coredns` package had some bits of
code that accessed/mutated pod annotations assuming that they'll always
be instantiated correctly.

This patch adds utility functions to safely mutate and access fargate
pod annotations.

Signed-off-by: Amine Hilaly <hilalyamine@gmail.com>

* Fix generating presigned URL for K8s authentication

With `aws-sdk-go-v2@1.24.1`, API server requests containing URLs presigned by `sts.PresignClient` fail with an `Unauthorized` error.

`aws-sdk-go-v2@1.24.1` adds an extra header `amz-sdk-request` to the generated request, but this header is not allow-listed by `aws-iam-authenticator` server running on the control plane.
This is likely due to [this change](aws/aws-sdk-go-v2#2438) which reorders the middleware operations to execute `RetryMetricsHeader` before `Signing`.

This changelist removes the `RetryMetricsHeader` middleware from the stack when constructing `sts.PresignClient`.

* Add release notes for 0.168.0

* Prepare for next development iteration

* Handle unordered public endpoint CIDRs from EKS in endpoint updates

For some clusters, EKS can return the list of public endpoint CIDRs out of
order, and won't allow updates where the incoming and current sets have set
equality (i.e. regardless of order of CIDR entries). This change restores the
set equality check that was removed in commit
72605fb and adds an additional test case to
cover this case.

* Fix outdated links

* Fix StringLike condition key for ebsCSIController IAM policy

The IAM condition key StringLike was used incorrectly in the policy and it doesn't work with wildcard (*) in the key itself. Wildcard is only supported in the value of the key. This fixes issue in cases where a volume dynamically provisioned via the older in-tree CSI plugin is being deleted by the new EBS CSI driver, because such volumes don't have the tags used in the policy.

The changes made are inspired from the AWS managed AmazonEBSCSIDriverPolicy.

* Fix coredns pdb preventing cluster deletion

* Add support for EKS 1.29

* Add release notes for 0.169.0

* Prepare for next development iteration

* Update arm-support.md

* Expand Karpenter settings.aws block to settings for v0.33.0 and greater

* Update stale.yml

* docs(fargate): eksctl update command is deprecated

* Fix deleting cluster sometimes drain managed nodegroups

* Update userdocs on nodegroups

* Update release drafter template

* Add support for Access Entry Type

* Update access entries userdocs

* Add release notes for 0.170.0

* Prepare for next development iteration

* Improve userdocs layout

* Announce eksctl Support Status Update

* Add support for Ubuntu 22.04 based EKS images

Ubuntu switched to 22.04 (Jammy) based images for EKS >= 1.29 . Add
support for that here.

* Add release notes for 0.171.0

* Prepare for next development iteration

* Changed the error to more understandable when the region code isn't set

* Fix checks for updated addon versions

When executing `eksctl get addons` an available update is not shown
if the only change to the version string is the `eksbuild` number, e.g.
`v1.0.0-eksbuild.2` is not shown as an update of `v1.0.0-eksbuild.1`.

The `findNewerVersions` func in `pkg/actions/get.go` had code to
explicitly ignore anything after the patch number. So `v1.1.0-eksbuild.1`
and `v1.1.0-eksbuild.4` were both converted to `1.1.0` and considered
equal. This fix removes the code that explicitly ignores the
`-eksbuild.x` portion of the version and allows the `semver` package
to compare the full version information.

Tests have also been updated to better match the version strings used
by AWS addons.

* Extract common steps setting up build environment from workflows

* Disable slack notifications

* Bump actions/cache from 3.3.2 to 4.0.0

* Update userdocs dependencies

* Update schema for new AMI family

* Update go dependencies

* Bump k8s dependencies to v0.29

Signed-off-by: Amine Hilaly <hilalyamine@gmail.com>

* Replace all deprecated sets.String with generic sets.Set

* Update actions dependencies

Update go version to 1.21 in actions

* Update build image manifest, tag file and workflows

* Update build image go version to 1.21

Update build image manifest, tag file and workflows

* Revert removing RetryMetricsHeader in presigned requests

* Add release notes for 0.172.0

* Revert misdeleted checkout step for publishing release

* Add integ test without build target

* Remove unused slack token

* Remove slackToken

* Prepare for next development iteration

* Bump dependencies for Dependabot alerts

Bump github.com/go-git/go-git/v5 from 5.4.2 to 5.11.0

Bumps [github.com/go-git/go-git/v5](https://github.com/go-git/go-git) from 5.4.2 to 5.11.0.
- [Release notes](https://github.com/go-git/go-git/releases)
- [Commits](go-git/go-git@v5.4.2...v5.11.0)

---
updated-dependencies:
- dependency-name: github.com/go-git/go-git/v5
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>

Bump helm.sh/helm/v3 from 3.14.0 to 3.14.2

Bumps [helm.sh/helm/v3](https://github.com/helm/helm) from 3.14.0 to 3.14.2.
- [Release notes](https://github.com/helm/helm/releases)
- [Commits](helm/helm@v3.14.0...v3.14.2)

---
updated-dependencies:
- dependency-name: helm.sh/helm/v3
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <support@github.com>

* Fix generate-internal-groups.sh permission

* Update dependencies

* Make EKS 1.29 default

* Automate PR with release notes draft

* Remove inline ELB and CloudWatch policies from Cluster Role (eksctl-io#7603)

* remove inline elb and cloud watch policy eksctl-io#7139

* fixed unit tests for removing extra inline policies

* Update docs with Ubuntu 20.04 supported in EKS <= 1.29 (eksctl-io#7618)

docs: ubuntu 20.04 supported in EKS <= 1.29

This adds a note that ubuntu focal (20.04) is supported up to EKS 1.29

* Bump github.com/cloudflare/circl from 1.3.3 to 1.3.7 (eksctl-io#7591)

Bumps [github.com/cloudflare/circl](https://github.com/cloudflare/circl) from 1.3.3 to 1.3.7.
- [Release notes](https://github.com/cloudflare/circl/releases)
- [Commits](cloudflare/circl@v1.3.3...v1.3.7)

---
updated-dependencies:
- dependency-name: github.com/cloudflare/circl
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Add profile flag support for get and delete access entry

* Fix caching credentials with assummed role MFA session

* Remove dependabot from contributors

* Use eksctl-bot token for release draft PR (eksctl-io#7629)

* Add release notes for v0.173.0

* Prepare for next development iteration

* Bump peter-evans/create-pull-request to fix error in GH API

* Update dependabot version updates frequency

* Fix EFA network interface device index assignment

* Remove goreleaser from build deps

Update build image manifest, tag file and workflows

* Replace make build-all with goreleaser action

* Inline script files in publish release workflows

* Rename release workflows

* Add release notes for v0.174.0

* Fix typo in publish-release.yaml

* Prepare for next development iteration

* Upgrade with explicit version if release version is up-to-date

* Test Bottlerocket node upgrade and verify version

* Add release notes for v0.175.0 (eksctl-io#7669)

* Add release notes for v0.175.0

* remove empty acknowledgements section

---------

Co-authored-by: yuxiang-zhang <23327251+yuxiang-zhang@users.noreply.github.com>
Co-authored-by: Tibi <110664232+TiberiuGC@users.noreply.github.com>

* Prepare for next development iteration (eksctl-io#7671)

* Bump dependencies (eksctl-io#7668)

* bump dependencies

* update mocks

* fix lint

* bump helm

* Aim for namespace uniqueness across parallel specs (eksctl-io#7680)

ensure namespace uniqueness across parallel specs

* Include MixedInstancesPolicy LaunchTemplate for validation

* Allow GPU instance types for Windows nodes (eksctl-io#7681)

* allow GPU instance type for Windows nodes

* update unit test for case gpus:0

* Display full draft release notes in PR description (eksctl-io#7686)

Update release-drafter.yaml

* Bump mkdocs version (eksctl-io#7696)

bump mkdocs version

* Add support for AMIs based on AmazonLinux2023 (eksctl-io#7684)

* add support for AL2023 for EKS-managed and self-managed nodes

* ensure AL2023 only supports containerd

* add GPU related validations + small nits

* add support for upgrades

* add support for EFA

* improve validations

* fix lint and unit tests

* update docs

* add validation error for maxpods limitation

* add integration tests for al2023

* improve validation message

* [EKSCTL create cluster command] Authorise self-managed nodes via `aws-auth configmap` when EKS access entries are disabled (eksctl-io#7698)

* Disable access entry creation for self-managed nodes on clusters with CONFIG_MAP only

* fix logic for updating aws-auth configmap

* Enforce `authenticationMode:CONFIG_MAP` on Outposts (eksctl-io#7699)

Make authenticationMode:CONFIG_MAP default on Outposts

* Add release notes for v0.176.0 (eksctl-io#7672)

Co-authored-by: TiberiuGC <110664232+TiberiuGC@users.noreply.github.com>

* Prepare for next development iteration

* Bump dependencies

Closes eksctl-io#7694 eksctl-io#7693 eksctl-io#7692 eksctl-io#7691 eksctl-io#7690 eksctl-io#7689 eksctl-io#7688 eksctl-io#7687 eksctl-io#7679 eksctl-io#7678 eksctl-io#7676 eksctl-io#7673 eksctl-io#7581 eksctl-io#7579 eksctl-io#7577 eksctl-io#7576

* Update build image tag

* Bump dependencies

* Fix arn build logic to support different aws partitions

* Fix reusing instanceRoleARN for nodegroups authorized with access entries

This changelist changes the design of creating access entries for self-managed nodegroups that use a pre-existing instanceRoleARN by creating the access entry resource outside of the CloudFormation stack by making a separate call to the AWS API. When deleting such a nodegroup, it's the user's responsibility to also delete the corresponding access entry when no more nodegroups are associated with it. This is because eksctl cannot tell if an access entry resource is still in use by non-eksctl created self-managed nodegroups.

Self-managed nodegroups not using a pre-existing instanceRoleARN will continue to have the access entry resource in the CloudFormation stack, making delete nodegroup an atomic operation for most use cases.

Fixes eksctl-io#7502

* Add note about deleting nodegroups

* Add integration tests

* Fix cluster deletion in tests

* Allow nodegroup creation after a cluster subnet is deleted (eksctl-io#7714)

* Preserve eksctl commands correctness when user deletes subnets

* update error when subnet availability validation fails

* address PR comments

* Handle K8s service account lifecycle on `eksctl create/delete podidentityassociation` commands (eksctl-io#7706)

* Handle K8s service account lifecycle on eksctl create/delete podidentityassociations commands

* correct typo

Co-authored-by: Chetan Patwal <cPu1@users.noreply.github.com>

---------

Co-authored-by: Chetan Patwal <cPu1@users.noreply.github.com>

* Add support for Ubuntu Pro 22.04 based EKS images (eksctl-io#7711)

* feat: Add support for Ubuntu Pro 22.04 based EKS images

* update schema.json

* test: Add nodegroup with Ubuntu Pro 22.04

* fix integration test

---------

Co-authored-by: Tibi <110664232+TiberiuGC@users.noreply.github.com>

* Disable IMDSv1 in unowned integration tests

* include pre-releases as full releases when drafting release notes

* Add utils command to migrate `iamidentitymappings` to EKS access entries (eksctl-io#7710)

* Added migrate-to-access-entry cmd structure

* Fix Target Authentication  mode validation

* Added logic to get accessEntries and cmEntries from cluster

* Added logic to make unique list of configmap accessEntries, and stack creation logic

* Added UpdateAuthentication mode and aeEntries filter logic

* Add approve flag check

* Added functionality to remove awsauth after switch to API only

* Adds logic to fetch FullARN of path stripped IAMIdentityMappings

* Updates some info log text

* Adds test case and refactors code

* Removes comments

* Adds taskTree and address PR comments

* Refactors code and Adds exception handling for NoSuchEntityException

* Resolves go.mod and go.sum conflicts

* Doc update for migrate-to-access-entry feature

* Fixed minimum iam policies doc to add permission for iam:GetUser

* Updated access-entries doc at migrate-to-access-entry section

* Fixes failing Migrate To Access Entry Test & go.mod, go.sum

* Amends migrate to access entry documentation

* improve logs and simplify code logic

* add unit tests

* ensure target-auth-mode has a valid value

---------

Co-authored-by: Pankaj Walke <advaitt@amazon.com>
Co-authored-by: Venkat Penmetsa <vpenmets@amazon.com>
Co-authored-by: Venkat Penmetsa <vpenmets@gmail.com>
Co-authored-by: Tibi <110664232+TiberiuGC@users.noreply.github.com>

* Revert "[Release drafter] Treat RCs as full releases when drafting notes" (eksctl-io#7725)

* Fix creating pod identities

Replaces usage of a per-loop variable with a per-iteration variable.

* Fix deleting pod identities

* Fix deleting clusters with a non-active status

* Add release notes for v0.177.0

* update release notes for 0.177.0

* change purchase type to capacity block when using capacity reservation

---------

Signed-off-by: Amine Hilaly <hilalyamine@gmail.com>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: Amine Hilaly <hilalyamine@gmail.com>
Co-authored-by: cpu1 <chetan@weave.works>
Co-authored-by: Yu Xiang Z <yxz.get@gmail.com>
Co-authored-by: Yu Xiang Zhang <yuxz@amazon.com>
Co-authored-by: eksctl-bot <53547694+eksctl-bot@users.noreply.github.com>
Co-authored-by: Nathaniel Emerson <nathaniel.emerson@skyscanner.net>
Co-authored-by: guessi <guessi@gmail.com>
Co-authored-by: Raghav Khandelwal <raghavk@zendrive.com>
Co-authored-by: Stephen Lang <stephen.lang@grafana.com>
Co-authored-by: Thomas Bechtold <thomas.bechtold@canonical.com>
Co-authored-by: Shuntaro Azuma <azush@amazon.com>
Co-authored-by: Matthew Robinson <matt@zensunni.org>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Osama Bin Junaid <32925504+ibnjunaid@users.noreply.github.com>
Co-authored-by: rpocase <rpocase@gmail.com>
Co-authored-by: Denys Havrysh <denys.gavrysh@gmail.com>
Co-authored-by: yuxiang-zhang <23327251+yuxiang-zhang@users.noreply.github.com>
Co-authored-by: Tibi <110664232+TiberiuGC@users.noreply.github.com>
Co-authored-by: Weifeng Wang <qclaogui@gmail.com>
Co-authored-by: Chetan Patwal <cPu1@users.noreply.github.com>
Co-authored-by: cpu1 <patwal.chetan@gmail.com>
Co-authored-by: TimAndy <xuchonglei@126.com>
Co-authored-by: cPu1 <patwal.chetan@gmail.comm>
Co-authored-by: Alberto Contreras <aciba90@gmail.com>
Co-authored-by: punkwalker <126026317+punkwalker@users.noreply.github.com>
Co-authored-by: Pankaj Walke <advaitt@amazon.com>
Co-authored-by: Venkat Penmetsa <vpenmets@amazon.com>
Co-authored-by: Venkat Penmetsa <vpenmets@gmail.com>
Co-authored-by: eV <evq@brave.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Issue in creating self-manged node when authenticationMode is set to CONFIG_MAP only [Bug]
3 participants