Light Dark Auto

Operator

Changelog for the greymatter.io Operator

0.10.0 (September 12, 2022)

Added

  • Add Bugsnag error reporting in the debug build.

  • Configurable operator namespace

  • Support multiple operators & meshes in the same K8s cluster

Changed

  • Removed the Mesh CRD and all dependent code and config

  • Removed the rest of the webhooks, and replaced sidecar injection webhooks with a controller-runtime Reconciliation loop and Finalizers

  • Sidecar injection annotations moved back up to the root metadata block of Deployemnts and StatefulSets.

  • Separate spire and deploy_spire flags and a defaults.spire block of configuration to optionally use a pre-existing Spire deployment

Fixed

  • Don't clobber existing cluster or workload labels in spec.template.metadata.labels

  • Unused vestigial mesh.spec.release_version is no longer required, and omitted by default

0.9.3 (August 11, 2022)

Added

  • Added support for using a tag as a GitOps target with -tag.

0.9.2 (July 15, 2022)

Changed

  • Upgraded the packaged greymatter.io CLI version to 4.2.

  • Updated submodules

0.9.1 (June 16, 2022)

Added

  • Watched namespaces that do not already exist are created before the imagePullSecret for retrieving the greymatter.io sidecar is copied in.

Changed

  • The annotations for deployment assist now need to be put in spec.template.metadata.annotations, not the top-level metadata.annotations for the Deployment or StatefulSet.

Fixed

  • A sidecar is no longer injected if the inject-sidecar-to annotation is missing from a Deployment or StatefulSet in a watched namespace.

0.9.0 (June 13, 2022)

Added

  • Added an annotation to control whether sidecar configuration is created for a given Deployment or StatefulSet, e.g., greymatter.io/configure-sidecar: "false"

  • Add a debug container build mode and instructions

  • OpenShift support restored, with automatic permissioning for core greymatter.io components

Changed

  • Completely new CUE for both Kubernetes manifests and greymatter.io config, rooted under pkg/cuemodule/core (inputs.cue, and gm/ and k8s/ subdirectories) following a philosophy of readability and malleability.

  • Operator Go now loads CUE generically, such that nearly all changes to deployment configuration require no changes to the Go.

  • Launch the operator with CUE-generated manifests, which also include the Spire manifests if the spire config flag is set.

  • Deployment assist is now opt-in, with labels "greymatter.io/inject-sidecar-to": "<port num>" (default is to omit the label) and "greymatter.io/configure-sidecar": "<boolean>" (default "true".)

  • Operator itself now runs as a StatefulSet rather than a Deployment.

  • One-Mesh/One-Operator - Multiple meshes require multiple operators (they do yet not play nicely together, but the target for multi-mesh support is to support graceful interop.)

  • Added a flag (auto_apply_mesh) that causes the operator to automatically deploy the default mesh defined in the CUE, after a delay. Turn the flag off to manually deploy a Mesh CR.

  • Switch to ECDSA 256 certificates for speed and security.

  • Checked in pkg/cuemodule/core/cue.mod/gen/k8s.io/, the result of cue get go k8s.io/api/..., for the CI build.

  • GitOps workflow now fetches the operator's CUE from a repository (default greymatter-io/greymatter-core) and polls it for changes which then trigger live updates to the Kubernetes and greymatter.io config of the live environment.

Fixed

  • Fixed parsing of the watch_namespaces field in the Mesh custom resource spec that prevented Control's GM_CONTROL_KUBERNETES_NAMESPACES environment variable from being set, causing service discovery to fail to identify data planes in the install_namespace when no watch_namespaces were specified.

  • Cleaned up unnecessary SPIRE agent unix workload attestor configuration.

  • Fixed family of issues that prevented the operator from restarting cleanly. The Operator now reloads the list of sidecars used for Spire mTLS authorization from the Mesh CR itself.

  • Fixed control-api URL in dashboard so that proxy config inspection and business impact selection would work from the UI.

0.3.3 (February 10, 2022)

This release includes internal changes to how we source greymatter.io mesh configuration schema. greymatter.io mesh configurations are defined in CUE and bundled into this project via a Git submodule greymatter-cue. Mesh custom resources are also ingested into CUE to be parsed for installing and configuring mesh deployments.

Additionally, this release addressed a bug that prevented the operator from restoring its own internal state of existing Mesh custom resources on startup. The fix ensures that the leader replica of the operator will always have the latest state of Mesh custom resources.

This release also has been tested for local development on Rancher Desktop.

Added

  • greymatter-cue Git submodule for importing greymatter.io CUE schemas.

  • Generated CUE schema for the Mesh custom resource definition, to be parsed within CUE when generating installation manifests and mesh configurations for a configured Mesh.

  • Fix concurrent write to a map shared among goroutines.

Changed

  • Moved all CUE packages into a single CUE module that lives in pkg/cuemodule/core. Each CUE package may be evaluated using the CUE CLI as well as loaded via exported loader functions defined in the cuemodule Go package.

  • Updated versions of Spire images (spire-server, spire-agent, and the k8s-workload-registrar) that the operator installs to 1.2.0. This fixes an issue on some platforms where Spire would misidentify the process as non-k8s, and fail to issue an identity.

  • Specifying --platform linux/amd64 in docker and buildah container builds to ensure M1 Macs know what to do with it, regardless of where it was built.

  • Increased memory limit (100Mi -> 300Mi) and increased initial wait period on readiness and liveness probes to accommodate slower/emulated machines (-> 120 sec).

Fixed

  • Updated flag in spire server liveness probe to prevent an eventual crash of spire server.

  • Addressed a bug that kept existing Mesh custom resources from being fully registered with a new instance of the operator on startup. The bug prevented new mesh configurations from being generated for workloads within namespaces of existing meshes.

0.3.0 (January 10, 2022)

Added

  • SPIRE as SPIFFE implementation for mutual TLS between workloads

Fixed

  • Prevent greymatter apply commands to Control API from running until it has connected to Redis.

Removed

  • Support for an external Redis cache for mesh configurations, opting for an internally-managed one secured by mutual TLS (via SPIRE).

  • Catalog entry for Redis

  • JWT Security service's Redis dependency

  • Listeners using port 10707 for initial boootstrapping of mesh configuration

0.2.0 (December 10, 2021)

Changed

  • Upgrade greymatter.io CLI binary dependency from 3.0.0 to 4.0.1

0.1.2 (December 2, 2021)

Changed

  • Change GREYMATTER_DOCKER_* config env vars to GREYMATTER_REGISTRY_*

  • Change k8s-operator --username and --password flags to --registry-username and --registry-password, respectively. And remove -u and -p aliases to reserve these for future use.

0.1.0 (December 1, 2021)

This is a pre-release with basic support for installing greymatter.io core components and dependencies and bootstrapping greymatter.io mesh configurations.

Added

  • Support for general Kubernetes distributions

  • Support for OpenShift, packaged for compatibility with the Operator Lifecycle Manager