Hi All,
Running into an issue with Flink Autoscaling
Implementation :
• Using Flink 1.20.1 with Flink Operator 1.12.0 for deployment
Issue :
• The job stops processing data in between during autoscaling whenever it scales up
• We want to auto-scale with Exactly Once and without Exactly Once jobs
What we tried :
• Enabled Unalligned checkpoints -> though the job never stopped processing, this resulted in data loss
Appreciate any inputs here.
Is there no way to ensure autoscaling without data loss and exactly-once semantics ?
CC: @Abhishek Joshi
Thanks.