v0.2

Kruize Version 0.2 released!

You can get it from
quay.io/kruize/autotune_operator:0.2
docker.io/kruize/autotune_operator:0.2
Branch: remote_monitoring, Tag: v0.2
Commit id: 268063a

Major changes include

[Deploy] Make local_monitoring as default
[API] Add experiment_name format to support ROS requirement
[API] Bulk API JSON format change to make it easier to parse experiment status
[API] Bulk API Add time range filter support to ROS input JSON
[API] Webhook support for Bulk API
[API] Datasource authentication DB updates
[Bug] Reduce default resource configuration for deploying Kruize onto OpenShift
[Bug] Upgrade OS to UBI Minimal 9.5 and other CVE fixes
[Bug] Better bulk API error handling
[Bug] promql query bug fixes
[Bug] Fix datasource exception management
[Demo] Better Kruize UX for local monitoring demo
[Tests] Bulk API functional tests

Full Changelog: v0.1...v0.2

v0.1

Kruize Version 0.1 released!

You can get it from
quay.io/kruize/autotune_operator:0.1
docker.io/kruize/autotune_operator:0.1
Branch: master, Tag: v0.1
Commit id: 02886da

This is a more experimental release than usual and it adds

1. Bulk API support
2. GPU Recommendations for MIG Partitioning
3. Make local_monitoring as default
4. Test updates for all of the above
5. Other bug fixes

Full Changelog: v0.0.25...v0.1

v0.0.25

Kruize Version 0.0.25_rm released!

You can get it from
quay.io/kruize/autotune_operator:0.0.25_rm
docker.io/kruize/autotune_operator:0.0.25_rm
Branch: remote_monitoring, Tag: v0.0.25
Commit id: 96be0e5

What's Changed

[Deploy] NA
[API] Support for Namespace recommendations (alpha)
[API] Add Queries for Namespace recommendations
[API] MetricsProfile updates
[Bug] Update OS, JVM and Maven versions
[Bug] Create metricProfile based on non-recording rules
[Tests] Fault tolerant test for kruize-local
[Tests] MetricProfile hardening tests

Full Changelog: v0.0.24...v0.0.25

v0.0.24

Kruize Version 0.0.24_rm released!

You can get it from
docker.io/kruize/autotune_operator:0.0.24_rm
quay.io/kruize/autotune_operator:0.0.24_rm
Branch: remote_monitoring, Tag: v0.0.24
Commit id: fdfb4b1

Release Changelog
[Deploy] Added 2 new kubernetes cluster type support AKS (Azure Kubernetes Service) and KinD
[API] Support for create, list and delete operations for metric profile
[API] Experimental Support for new Kubernetes workload type ‘job’
[Bug] Resolved the Kruize prometheus notification issue
[Tests] End to end local monitoring testing
[Tests] Tests to support Metric Profile API operations

Full list of PRs can be found here

v0.0.20

Kruize Version 0.0.20_rm released!

You can get it from
docker pull docker.io/kruize/autotune_operator:0.0.20_rm
docker pull quay.io/kruize/autotune_operator:0.0.20_rm
Commit id: da75ffe66d1677375167a9c88046b0dbaadf41ac

Release Changelog

[API] V2 of the API, updates the listRecommendations / updateRecommendations API JSON to add the new recommendations profile.
[Deploy] No updates
[UI] A new version of the Kruize UI (0.0.3) has been released that supports the new recommendation profiles.
[DB] No updates
[Workflow] No updates
[Recommendations] update_recommendations JSON has been updated to add “cost” and “performance” recommendation engines.
[Recommendations] The definition of the two recommendation engines (or profiles) is documented here.
[Test] Test automation improvements
[Test] Running tests through comments
[Test] Script to grab Kruize metrics from prometheus
[Bug] Better validation of updtes_results params
Full list of PRs can be found here.

The primary focus of this release is adding the recommendations profile. We now support two different recommendation “profiles”, cost and performance. Currently they only affect CPU recommendations alone. Cost recommendations correspond to the 60th percentile of the CPU usage, while the performance profile recommendations correspond to the 98th percentile of CPU usage for the related container.