Use nonroot base images#8408
Conversation
- Use nonroot-<arch> tags for STATIC_BASE image digests - Use nonroot-<arch> tags for DYNAMIC_BASE image digests - Update hack/latest-base-images.sh to emit nonroot digests Signed-off-by: Richard Wall <richard.wall@cyberark.com>
Signed-off-by: Richard Wall <richard.wall@cyberark.com>
Instead inherit the default user and group from the nonroot base images Signed-off-by: Richard Wall <richard.wall@cyberark.com>
|
Maybe we should start using |
Maybe, but that's a different problem and can be addressed in another PR. Switching from a Debian base image to an Alpine / APKO base image needs further discussion, I think. |
There was a problem hiding this comment.
Pull request overview
This PR migrates cert-manager container images from custom UID 1000 (with GID 0/root) to using distroless nonroot base images with UID/GID 65532. The change follows container security best practices by ensuring both the user and group are non-root, addressing the previous configuration where containers ran as UID 1000 but with the root group (GID 0).
Changes:
- Updated base image references to use
:nonroot-*tags instead of:latest-*for all supported architectures - Removed explicit
USER 1000directives from all Containerfiles (controller, webhook, cainjector, acmesolver, startupapicheck) - Modified the
hack/latest-base-images.shscript to pull nonroot image variants
Reviewed changes
Copilot reviewed 7 out of 7 changed files in this pull request and generated no comments.
Show a summary per file
| File | Description |
|---|---|
| make/base_images.mk | Updated base image digests to nonroot variants for both static and dynamic images across all architectures |
| hack/latest-base-images.sh | Modified to fetch nonroot image tags instead of latest tags, with added documentation |
| hack/containers/Containerfile.webhook | Removed explicit USER 1000 directive (inherited from base image) |
| hack/containers/Containerfile.startupapicheck | Removed explicit USER 1000 directive (inherited from base image) |
| hack/containers/Containerfile.controller | Removed explicit USER 1000 directive (inherited from base image) |
| hack/containers/Containerfile.cainjector | Removed explicit USER 1000 directive (inherited from base image) |
| hack/containers/Containerfile.acmesolver | Removed explicit USER 1000 directive (inherited from base image) |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
|
I agree with moving to 65532:65532! /approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: inteon The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
📢 The fix or feature is now available for testing in an alpha release Please test and report back. |
…#4581) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [cert-manager/cert-manager](https://github.com/cert-manager/cert-manager) | minor | `v1.19.4` → `v1.20.0` | --- ### Release Notes <details> <summary>cert-manager/cert-manager (cert-manager/cert-manager)</summary> ### [`v1.20.0`](https://github.com/cert-manager/cert-manager/releases/tag/v1.20.0) [Compare Source](cert-manager/cert-manager@v1.19.4...v1.20.0) cert-manager is the easiest way to automatically manage certificates in Kubernetes and OpenShift clusters. v1.20.0 adds support for the new ListenerSet resource, adds support for Azure Private DNS; parentRefs are no longer required when using ACME with Gateway API, and OtherNames was promoted to Beta. #### Changes by Kind ##### Feature - Added a set of flags to permit setting NetworkPolicy across all deployed containers. Remove redundant global IP ranges from example policies. ([#​8370](cert-manager/cert-manager#8370), [@​jcpunk](https://github.com/jcpunk)) - Added selectable fields to custom resource definitions for .spec.issuerRef.{group, kind, name} ([#​8256](cert-manager/cert-manager#8256), [@​tareksha](https://github.com/tareksha)) - Added support for specifying `imagePullSecrets` in the `startupapicheck-job` Helm template to enable pulling images from private registries. ([#​8186](cert-manager/cert-manager#8186), [@​mathieu-clnk](https://github.com/mathieu-clnk)) - Added 'extraContainers' helm chart value, allowing the deployment of arbitrary sidecar containers within the cert-manager operator pod. This can be used to support, for e.g., AWS IAM Roles Anywhere for Route53 DNS01 verification. ([#​8355](cert-manager/cert-manager#8355), [@​dancmeyers](https://github.com/dancmeyers)) - Added `parentRef` override annotations on the Certificate resource. ([#​8518](cert-manager/cert-manager#8518), [@​hjoshi123](https://github.com/hjoshi123)) - Added support for azure private zones for dns01 issuer. ([#​8494](cert-manager/cert-manager#8494), [@​hjoshi123](https://github.com/hjoshi123)) - Added support for configuring PEM decoding size limits, allowing operators to handle larger certificates and keys. ([#​7642](cert-manager/cert-manager#7642), [@​robertlestak](https://github.com/robertlestak)) - Added support for unhealthyPodEvictionPolicy in PodDisruptionBudget ([#​7728](cert-manager/cert-manager#7728), [@​jcpunk](https://github.com/jcpunk)) - For Venafi provider, read `venafi.cert-manager.io/custom-fields` annotation on Issuer/ClusterIssuer and use it as base with override/append capabilities on Certificate level. ([#​8301](cert-manager/cert-manager#8301), [@​k0da](https://github.com/k0da)) - Improve error message when CA issuers are misconfigured to use a clashing secret name ([#​8374](cert-manager/cert-manager#8374), [@​majiayu000](https://github.com/majiayu000)) - Introduce a new Ingress annotation `acme.cert-manager.io/http01-ingress-ingressclassname` to override `http01.ingress.ingressClassName` field in HTTP-01 challenge solvers. ([#​8244](cert-manager/cert-manager#8244), [@​lunarwhite](https://github.com/lunarwhite)) - Update `global.nodeSelector` to helm chart to perform a `merge` and allow for a single `nodeSelector` to be set across all services. ([#​8195](cert-manager/cert-manager#8195), [@​StingRayZA](https://github.com/StingRayZA)) - Vault issuers will now include the Vault server address as one of the default audiences on generated service account tokens. ([#​8228](cert-manager/cert-manager#8228), [@​terinjokes](https://github.com/terinjokes)) - Added experimental `XListenerSet` feature gate ([#​8394](cert-manager/cert-manager#8394), [@​hjoshi123](https://github.com/hjoshi123)) ##### Documentation - Add GWAPI documentation to NOTES.TXT in helm chart ([#​8353](cert-manager/cert-manager#8353), [@​jaxels10](https://github.com/jaxels10)) ##### Bug or Regression - Adds logs for cases when acme server returns us a fatal error in the order controller ([#​8199](cert-manager/cert-manager#8199), [@​Peac36](https://github.com/Peac36)) - Fixed an issue where kind or group in the issuerRef of a Certificate was omitted, upgrading to 1.19.x incorrectly caused the certificate to be renewed ([#​8160](cert-manager/cert-manager#8160), [@​inteon](https://github.com/inteon)) - Changes to the Duration and RenewBefore annotations on ingress and gateway-api resources will now trigger certificate updates. ([#​8232](cert-manager/cert-manager#8232), [@​eleanor-merry](https://github.com/eleanor-merry)) - Fix an issue where ACME challenge TXT records are not cleaned up when there are many resource records in CloudDNS. ([#​8456](cert-manager/cert-manager#8456), [@​tkna](https://github.com/tkna)) - Fix unregulated retries with the DigitalOcean DNS-01 solver Add full detailed DNS-01 errors to the events attached to the Challenge, for easier debugging ([#​8221](cert-manager/cert-manager#8221), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Fixed an infinite re-issuance loop that could occur when an issuer returns a certificate with a public key that doesn't match the CSR. The issuing controller now validates the certificate before storing it and fails with backoff on mismatch. ([#​8403](cert-manager/cert-manager#8403), [@​calm329](https://github.com/calm329)) - Fixed an issue where HTTP-01 challenges failed when the Host header contains an IPv6 address. This means that users can now issue IP address certificates for IPv6 address subjects. ([#​8424](cert-manager/cert-manager#8424), [@​SlashNephy](https://github.com/SlashNephy)) - Fixed the HTTP-01 Gateway solver creating invalid HTTPRoutes by not setting spec.hostnames when the challenge DNSName is an IP address. ([#​8443](cert-manager/cert-manager#8443), [@​alviss7](https://github.com/alviss7)) - Revert API defaults for issuer reference kind and group introduced in 0.19.0 ([#​8173](cert-manager/cert-manager#8173), [@​erikgb](https://github.com/erikgb)) - Security (MODERATE): Fix a potential panic in the cert-manager controller when a DNS response in an unexpected order was cached. If an attacker was able to modify DNS responses (or if they controlled the DNS server) it was possible to cause denial of service for the cert-manager controller. ([#​8469](cert-manager/cert-manager#8469), [@​SgtCoDFish](https://github.com/SgtCoDFish)) - Update Go to `v1.25.5` to fix `CVE-2025-61727` and `CVE-2025-61729` ([#​8290](cert-manager/cert-manager#8290), [@​octo-sts](https://github.com/octo-sts)\[bot]) - When Prometheus monitoring is enabled, the metrics label is now set to the intended value of `cert-manager`. Previously, it was set depending on various factors (namespace cert-manager is installed in and/or Helm release name). ([#​8162](cert-manager/cert-manager#8162), [@​LiquidPL](https://github.com/LiquidPL)) ##### Other (Cleanup or Flake) - Promoted the OtherNames feature to Beta and enabled it by default ([#​8288](cert-manager/cert-manager#8288), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Promoting `xlistenerset` feature gate to `listenerset` ([#​8501](cert-manager/cert-manager#8501), [@​hjoshi123](https://github.com/hjoshi123)) - Rebranding of the Venafi Issuer to CyberArk ([#​8215](cert-manager/cert-manager#8215), [@​iossifbenbassat123](https://github.com/iossifbenbassat123)) - Switched to SSA for challenge finalizer updates ([#​8519](cert-manager/cert-manager#8519), [@​inteon](https://github.com/inteon)) - The default container user (UID) is now 65532 (previously 1000) and the default container group (GID) is now 65532 (previously 0) ([#​8408](cert-manager/cert-manager#8408), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - The feature-gate DefaultPrivateKeyRotationPolicyAlways moved from Beta to GA and can no longer be disabled. ([#​8287](cert-manager/cert-manager#8287), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Update cert-manager's ACME client, forked from golang/x/crypto ([#​8268](cert-manager/cert-manager#8268), [@​SgtCoDFish](https://github.com/SgtCoDFish)) - Use the latest version of Kyverno (1.16.2) in the best-practice installation tests ([#​8389](cert-manager/cert-manager#8389), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - We stopped testing with Coutour due to it not supporting the new XListenerSet resource, and moved to kgateway. ([#​8426](cert-manager/cert-manager#8426), [@​hjoshi123](https://github.com/hjoshi123)) </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box --- This PR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My41OS4yIiwidXBkYXRlZEluVmVyIjoiNDMuNTkuMiIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOlsiaW1hZ2UiXX0=--> Reviewed-on: https://gitea.alexlebens.dev/alexlebens/infrastructure/pulls/4581 Co-authored-by: Renovate Bot <renovate-bot@alexlebens.net> Co-committed-by: Renovate Bot <renovate-bot@alexlebens.net>
This PR contains the following updates: | Package | Update | Change | |---|---|---| | [cert-manager](https://cert-manager.io) ([source](https://github.com/cert-manager/cert-manager)) | minor | `v1.19.4` → `v1.20.0` | --- ### Release Notes <details> <summary>cert-manager/cert-manager (cert-manager)</summary> ### [`v1.20.0`](https://github.com/cert-manager/cert-manager/releases/tag/v1.20.0) [Compare Source](cert-manager/cert-manager@v1.19.4...v1.20.0) cert-manager is the easiest way to automatically manage certificates in Kubernetes and OpenShift clusters. v1.20.0 adds support for the new ListenerSet resource, adds support for Azure Private DNS; parentRefs are no longer required when using ACME with Gateway API, and OtherNames was promoted to Beta. #### Changes by Kind ##### Feature - Added a set of flags to permit setting NetworkPolicy across all deployed containers. Remove redundant global IP ranges from example policies. ([#​8370](cert-manager/cert-manager#8370), [@​jcpunk](https://github.com/jcpunk)) - Added selectable fields to custom resource definitions for .spec.issuerRef.{group, kind, name} ([#​8256](cert-manager/cert-manager#8256), [@​tareksha](https://github.com/tareksha)) - Added support for specifying `imagePullSecrets` in the `startupapicheck-job` Helm template to enable pulling images from private registries. ([#​8186](cert-manager/cert-manager#8186), [@​mathieu-clnk](https://github.com/mathieu-clnk)) - Added 'extraContainers' helm chart value, allowing the deployment of arbitrary sidecar containers within the cert-manager operator pod. This can be used to support, for e.g., AWS IAM Roles Anywhere for Route53 DNS01 verification. ([#​8355](cert-manager/cert-manager#8355), [@​dancmeyers](https://github.com/dancmeyers)) - Added `parentRef` override annotations on the Certificate resource. ([#​8518](cert-manager/cert-manager#8518), [@​hjoshi123](https://github.com/hjoshi123)) - Added support for azure private zones for dns01 issuer. ([#​8494](cert-manager/cert-manager#8494), [@​hjoshi123](https://github.com/hjoshi123)) - Added support for configuring PEM decoding size limits, allowing operators to handle larger certificates and keys. ([#​7642](cert-manager/cert-manager#7642), [@​robertlestak](https://github.com/robertlestak)) - Added support for unhealthyPodEvictionPolicy in PodDisruptionBudget ([#​7728](cert-manager/cert-manager#7728), [@​jcpunk](https://github.com/jcpunk)) - For Venafi provider, read `venafi.cert-manager.io/custom-fields` annotation on Issuer/ClusterIssuer and use it as base with override/append capabilities on Certificate level. ([#​8301](cert-manager/cert-manager#8301), [@​k0da](https://github.com/k0da)) - Improve error message when CA issuers are misconfigured to use a clashing secret name ([#​8374](cert-manager/cert-manager#8374), [@​majiayu000](https://github.com/majiayu000)) - Introduce a new Ingress annotation `acme.cert-manager.io/http01-ingress-ingressclassname` to override `http01.ingress.ingressClassName` field in HTTP-01 challenge solvers. ([#​8244](cert-manager/cert-manager#8244), [@​lunarwhite](https://github.com/lunarwhite)) - Update `global.nodeSelector` to helm chart to perform a `merge` and allow for a single `nodeSelector` to be set across all services. ([#​8195](cert-manager/cert-manager#8195), [@​StingRayZA](https://github.com/StingRayZA)) - Vault issuers will now include the Vault server address as one of the default audiences on generated service account tokens. ([#​8228](cert-manager/cert-manager#8228), [@​terinjokes](https://github.com/terinjokes)) - Added experimental `XListenerSet` feature gate ([#​8394](cert-manager/cert-manager#8394), [@​hjoshi123](https://github.com/hjoshi123)) ##### Documentation - Add GWAPI documentation to NOTES.TXT in helm chart ([#​8353](cert-manager/cert-manager#8353), [@​jaxels10](https://github.com/jaxels10)) ##### Bug or Regression - Adds logs for cases when acme server returns us a fatal error in the order controller ([#​8199](cert-manager/cert-manager#8199), [@​Peac36](https://github.com/Peac36)) - Fixed an issue where kind or group in the issuerRef of a Certificate was omitted, upgrading to 1.19.x incorrectly caused the certificate to be renewed ([#​8160](cert-manager/cert-manager#8160), [@​inteon](https://github.com/inteon)) - Changes to the Duration and RenewBefore annotations on ingress and gateway-api resources will now trigger certificate updates. ([#​8232](cert-manager/cert-manager#8232), [@​eleanor-merry](https://github.com/eleanor-merry)) - Fix an issue where ACME challenge TXT records are not cleaned up when there are many resource records in CloudDNS. ([#​8456](cert-manager/cert-manager#8456), [@​tkna](https://github.com/tkna)) - Fix unregulated retries with the DigitalOcean DNS-01 solver Add full detailed DNS-01 errors to the events attached to the Challenge, for easier debugging ([#​8221](cert-manager/cert-manager#8221), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Fixed an infinite re-issuance loop that could occur when an issuer returns a certificate with a public key that doesn't match the CSR. The issuing controller now validates the certificate before storing it and fails with backoff on mismatch. ([#​8403](cert-manager/cert-manager#8403), [@​calm329](https://github.com/calm329)) - Fixed an issue where HTTP-01 challenges failed when the Host header contains an IPv6 address. This means that users can now issue IP address certificates for IPv6 address subjects. ([#​8424](cert-manager/cert-manager#8424), [@​SlashNephy](https://github.com/SlashNephy)) - Fixed the HTTP-01 Gateway solver creating invalid HTTPRoutes by not setting spec.hostnames when the challenge DNSName is an IP address. ([#​8443](cert-manager/cert-manager#8443), [@​alviss7](https://github.com/alviss7)) - Revert API defaults for issuer reference kind and group introduced in 0.19.0 ([#​8173](cert-manager/cert-manager#8173), [@​erikgb](https://github.com/erikgb)) - Security (MODERATE): Fix a potential panic in the cert-manager controller when a DNS response in an unexpected order was cached. If an attacker was able to modify DNS responses (or if they controlled the DNS server) it was possible to cause denial of service for the cert-manager controller. ([#​8469](cert-manager/cert-manager#8469), [@​SgtCoDFish](https://github.com/SgtCoDFish)) - Update Go to `v1.25.5` to fix `CVE-2025-61727` and `CVE-2025-61729` ([#​8290](cert-manager/cert-manager#8290), [@​octo-sts](https://github.com/octo-sts)\[bot]) - When Prometheus monitoring is enabled, the metrics label is now set to the intended value of `cert-manager`. Previously, it was set depending on various factors (namespace cert-manager is installed in and/or Helm release name). ([#​8162](cert-manager/cert-manager#8162), [@​LiquidPL](https://github.com/LiquidPL)) ##### Other (Cleanup or Flake) - Promoted the OtherNames feature to Beta and enabled it by default ([#​8288](cert-manager/cert-manager#8288), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Promoting `xlistenerset` feature gate to `listenerset` ([#​8501](cert-manager/cert-manager#8501), [@​hjoshi123](https://github.com/hjoshi123)) - Rebranding of the Venafi Issuer to CyberArk ([#​8215](cert-manager/cert-manager#8215), [@​iossifbenbassat123](https://github.com/iossifbenbassat123)) - Switched to SSA for challenge finalizer updates ([#​8519](cert-manager/cert-manager#8519), [@​inteon](https://github.com/inteon)) - The default container user (UID) is now 65532 (previously 1000) and the default container group (GID) is now 65532 (previously 0) ([#​8408](cert-manager/cert-manager#8408), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - The feature-gate DefaultPrivateKeyRotationPolicyAlways moved from Beta to GA and can no longer be disabled. ([#​8287](cert-manager/cert-manager#8287), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - Update cert-manager's ACME client, forked from golang/x/crypto ([#​8268](cert-manager/cert-manager#8268), [@​SgtCoDFish](https://github.com/SgtCoDFish)) - Use the latest version of Kyverno (1.16.2) in the best-practice installation tests ([#​8389](cert-manager/cert-manager#8389), [@​wallrj-cyberark](https://github.com/wallrj-cyberark)) - We stopped testing with Coutour due to it not supporting the new XListenerSet resource, and moved to kgateway. ([#​8426](cert-manager/cert-manager#8426), [@​hjoshi123](https://github.com/hjoshi123)) </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box --- This PR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate). <!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My41OS4yIiwidXBkYXRlZEluVmVyIjoiNDMuNTkuMiIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOlsiY2hhcnQiXX0=--> Reviewed-on: https://gitea.alexlebens.dev/alexlebens/infrastructure/pulls/4582 Co-authored-by: Renovate Bot <renovate-bot@alexlebens.net> Co-committed-by: Renovate Bot <renovate-bot@alexlebens.net>
The cert-manager containers were configured to run with UID 1000, but the group was unspecified which caused the cert-manager processes to have the root group (GID: 0) by default.
We don't think this is particularly scary from a security perspective because cert-manager's pods are so minimal that there's little that GID 0 gets you.
And by default, the Pods are configured with a security context which drops all privileges.
Nevertheless, it is best-practice to use a non-root group and it makes sense to use the same user and group as were chosen for the distroless base-images for the reasons given in:
This was brought to our attention by a user in Slack: https://kubernetes.slack.com/archives/C4NV3DWUC/p1768411371860499
/kind cleanup
Backwards compatibility
There's some discussion about switching distroless images to use non-root by default and it has been rejected because it risks breaking existing users, but I don't think any of those arguments applies to cert-manager which is designed to run as non-root.
Testing
Compile and deploy cert-manager in a Kind cluster
Check the Pod status for the actual UID and GID of the container.
Before:
After:
I also investigated whether we could test this using Kyverno configured to reject containers with GID 0, but that only works if we add a
runAsGroupto the pod templates, I think.Here's an example of such a policy, but it is not included among Kyverno's restricted pod security standards (against which we already test cert-manager for compliance):
CyberArk tracker: VC-48682