v0.2.1

Changelog

• Don't Fatal for NonExistent Queue #33 #34
• Bump release #35 #36 #38 #39 #40

v0.2.0

Changelog

#24 Ignore the error due to the api server log #25
#26 #27 Desired should not be incremented when the messages are below target.
#29 Prevent DLQ issue when bootup time is more; solved by #31

v0.1.3

Changelog

• #22 Retry request should not fatal on error
• #23 Race fixes, sync thread stoppage bug fixed
  • Race fixed on list poll threads
  • Sync thread was returning(stopping) on the dead thread cleanup. Fixed #14

v0.1.2

Changelog

• #17 #20 Fixes race conditions
  • Don't work with pointers across go-routines - mv *QueueSpec with QueueSpec to prevent race
  • Use a deep copy of a map across go-routines as maps are always passed by reference
  • Info formatted for queue status
• #18 Don't fatal for this issue and log as error. This was fixed in k8s 1.13.1

v0.1.1

Changelog

• #13 Fixes worker pod autoscaler crash when updating deployments by adding retry implmentation
• #15 Tuned worker pod autoscaler for the following changes
  • Bug fix to wait for short poll interval when appoxMessagesVisible > 0
  • Status of WPA should have current metric information
  • Don't fatal when queue is not found as it is a central controller
  • Desired replica calculation changed to prevent un-necessary scale ups #16

v0.1.0

Changelog

• #1 Init Project, vendor management, Makefile, build, format scripts, command line init code, init package and README defining briefly planned work.
• #2 #3 #5 CRD creations, Project made generic and renamed to worker pod autoscaler.
• #4 WPA API, generators, Core controller written following the standard set by sample-controller
• #6 Queue based autoscaling core code
  • Queue package
  • Controller package
  • Generic Poller as the thread manager for all the queues polling
  • QueuingService interface
  • SQS implementation of QueuingService interface
  • Major work on making every piece fast and concurrent using go channels
  • Scale up and down logic changes and test cases(play.golang.org)
• #7 Scale down related improvements
• #8 Cache empty receive sqs API call as it is refreshed only in 5 mins interval
• #9 Give control to control the speed and concurrency - Added following flags
  • --resync-period
  • --sqs-long-poll-interval
  • --sqs-short-poll-interval
• #10 #11 #12 Made workerpodautoscaler kubernetes ready. Build and deployment realated changes.