Releases: practo/k8s-worker-pod-autoscaler
Releases · practo/k8s-worker-pod-autoscaler
v1.6.0
Changelog
- #145 introduces a cooldown. Scale down delay of 10mins(default, configurable) after last scale activity. This prevents repeated scale down for workers whose WPA status are updated very frequently. This will make the scaling down smooth and prevent spiky scales ups and downs. Fixes #143. Thanks @michael-careplanner.
- #142 Updates CRD to use v1 for k8s 1.22+ support.
- #139 Fixes #53 slows down WPA for q's that does not exist.
Images
Images are available in Dockerhub and AWS public ECR:
practodev/workerpodautoscaler:v1.6.0
practodev/workerpodautoscaler:v1.6
practodev/workerpodautoscaler:v1
public.ecr.aws/practo/workerpodautoscaler:v1.6.0
public.ecr.aws/practo/workerpodautoscaler:v1.6
public.ecr.aws/practo/workerpodautoscaler:v1
Contributors
michael-careplanner
Assets 2
v1.5.0
Changelog
- #137 Fixes #136. WorkerPodAutoscaler pod on restarts was scaling workers to minReplicas. This was happening because it was taking scaling decisions based on uninitialized data. This was fixed, and now the control loop does not take scaling decisions if the data is not initliazed. Thanks @aleclerc-sonrai
- #138 Upgrades Golang to 1.17.1, updates kubernetes to 1.21.4 and also other dependencies.
Images are available in Dockerhub and AWS public ECR:
Images
practodev/workerpodautoscaler:v1.5.0
practodev/workerpodautoscaler:v1.5
practodev/workerpodautoscaler:v1
public.ecr.aws/practo/workerpodautoscaler:v1.5.0
public.ecr.aws/practo/workerpodautoscaler:v1.5
public.ecr.aws/practo/workerpodautoscaler:v1
Contributors
aleclerc-sonrai
Assets 2
v1.4.1
v1.4.0
Changelog
Changes
- #126 Kubernetes 1.20 support and Golang upgraded to 1.16.4
- WPA images are released to Public ECR from this release as well. Dockerhub images are still the default.
practodev/workerpoduatoscaler:v1
practodev/workerpoduatoscaler:v1.4
practodev/workerpoduatoscaler:v1.4.0
public.ecr.aws/practo/workerpodautoscaler:v1
public.ecr.aws/practo/workerpodautoscaler:v1.4
public.ecr.aws/practo/workerpodautoscaler:v1.4.0
v1.3.0
v1.2.0
Changelog
Changes
- #108 Add omitempty to optional spec types. @matkam
- #109 Verbosity level changes in the logging across the WPA code following the k8s standard.
- #112 Use sync map for storing caches in SQS. Fixed #110
- #114 Vendor updates
- #115 #118 Golang was updated. Current golang is: 1.15.3
- #120 Adds support for namespace scope in WPA.
--namespace. Fixed #59 completely now. @nqkdev - #119 Prevent crash due to connection failures in both SQS and Beanstalk.
- #123 Release changes for 1.2.0
v1.1.0
Changelog
Changes
- #101 #67 #105 #97 Fixes desired logic calculations @AmeerAssi @jotish @sujanadiga
- #98 Fixed a major API cache bug in #97 @AmeerAssi
- #96 Support for replica sets in WPA @matkam
- #89 Partial scale down was not happening when messages were below target. This was fixed in #90
- #74 Queue and Worker Metrics for WPA. Helps in visualising and analysing the scaling better. Went in #91
- #70 Adds a rule to keep calculated min(min based on RPM) below MAX. Fixed in #93 @ishanjoshi02
- #94 #95 Fixes the WPA installation script @keithnull
- #105 #103 Document update to clear env WPA_ confusions. @agconti
- #100 v1.1.0 Release changes
Significant changes that went out in v1.0.0 before v1.1.0
- #69 Beanstalk Support (includes #57). @sushant-115
- Following WPA specs were added making scaling reliable in v1.0.0-beta release.
secondsToProcessOneJob
maxDisruption
v1.0.0
Changelog
Significant changes that went out after beta
- #69 Beanstalk Support (includes #57). @sushant-115
- #84 #59 #68 #56 CRD creation moves out from code to YAML, version v1 added, CRD openAPIV3 validation added.
- #80 Partial scale down was not happening in one scenario. Fixed in #87.
- #73 Scale up logic does not work well for long-running tasks. Fixed in #86.
Significant changes that went out in beta
Following WPA specs were added making scaling reliable in v1.0.0-beta release.
secondsToProcessOneJob
maxDisruption
Changes
- #84 CRD and its validation added.
- #72 Error instead of fatal, and log infor of min and maxDisruptable
- #78 Add Verbosity level
- #77 #79 #81 #83 Prometheus metrics for the WPA logs using klog and promlog
wpa_log_messages_total{severity="ERROR"} 0
wpa_log_messages_total{severity="WARNING"} 0
- #85 Publish major, major-minor tags along with complete semver tags. @ryantate13
v1.0.0-beta
Changelog
secondsToProcessOneJob
maxDisruption
wpa_controller_loop_count_success{workerpodautoscaler="priority-notification-10d4e"} 11
wpa_controller_loop_duration_seconds{workerpodautoscaler="priority-notification-10d4e"} 159
go_goroutines{endpoint="workerpodautoscaler-metrics"} 5843
--k8s-api-qps
--k8s-api-burst
#62: PR did a lot of improvements. For details please check the PR description: