Releases: kubeflow/spark-operator
Release v1beta2-1.0.1 with Spark 2.4.4
Release v1beta2-1.0.0 with Spark 2.4.4
This is the first release with the new v1beta2
version of the API (#578). This release also includes an upgrade to Spark 2.4.4 (#625) for both the operator image and Spark images for the various language bindings. This release includes the following enhancements and bug fixes:
- Support for high-availability mode with leader election (#518, #575).
- Allow
PodSecurityContext
to be overridden (#528). - Spark pod mutating webhook enhancements (#543).
- A new
ResourceQuota
enforcement webhook (#544 ). - Switch to using
controller-gen
to generate/update CRD yamls (#578). - Integration with Volcano for batch scheduling (#567, #585, #599, #606).
- TTL support for
SparkApplication
s (#615). - Bug fixes.
- Documentation enhancements.
Release v1beta1-0.9.0
This release includes the following enhancements and fixes:
- Support for adding sidecar containers into the driver and executor pods (#478).
- Support for requesting GPUs (#509).
- Support for specifying hostNetwork for the driver and executor pods (#510).
- Support for specifying node selectors at pod level (#504).
- Support for specifying DNS config options for the driver and executor pods (#519).
- Changed the way driver pod status is checked (#506). The operator now sets the name of the driver pod while submitting an application to run and uses the name to get the driver pod instead of listing the Spark pods and filter out the driver pod.
- Changed the type of the driver UI service from NodePort to ClusterIP (#520).
- Other fixes and documentation enhancements.
Release v1beta1-0.8.2
Merge pull request #477 from liyinan926/master Fixed mounting of Prometheus configuration ConfigMap
Release v1beta1-0.8.1
This release contains an import bug fix for an issue related to application spec update and restart handling. Specifically it contains:
Release v1beta1-0.8.0
This release includes the following changes:
- The mutation webhook now gets
SparkApplication
objects and use their specs to figure out what to mutate to Spark driver and executor Pods. This replaced the original approach of using special annotations to carry needed mutations. This makes it much easier to support future mutations. - Added support for specifying
PodSecurityContext
for the driver and/or executor Pods. - Pinned the Kubernetes dependencies to a specific version instead of a release branch.
- Other bug fixes.
Release v1beta1-0.7.1
This beta release includes the following changes:
- Various bug and documentation fixes.
- Better jsonschema validation.
Release v1beta1-0.7
This is the first beta release of the operator after switching to use the v1beta1
version of the APIs.
Release v1alpha1-0.6
This is an alpha release with the v1alpha1 version of the CRDs. This release includes the following changes:
- Fixes to allow running multiple instances of the operator with each one for a different namespace.
- Support for OpenShift clusters.
- Fixes to make the scheduling logic for
ScheduledSparkApplication
s more robust. - Other fixes and enhancements.
Release v1alpha1-0.5
This is an alpha release with the v1alpha1 version of the CRDs. This release includes the following changes:
- Upgraded to Spark 2.4.0.
- Major refactoring to the SparkApplication controller logic around application/driver/executor state tracking and management.
- Optional ingress creation for Spark driver UI.
- Basic end-to-end tests.
- Other bug fixes and enhancements.