tree 05e2f2682e668cadcb08b6e961c09d5a9aaa538e
parent d0ce3d855305bf602a558aee1049d7e37c86c895
author Tudor Pavel <tudorv.pavel@gmail.com> 1697809627 +0300
committer GitHub <noreply@github.com> 1697809627 -0700
gpgsig -----BEGIN PGP SIGNATURE-----
 
 wsBcBAABCAAQBQJlMoTbCRBK7hj4Ov3rIwAA7YcIABoLhBNXapHRcBcB6+p61kyG
 rX0qeOiiVG4+PQwtbfsIzVltEiJT7OXbfTLsKoImQqcOOC4O/UcL7wON59ngnF4G
 JXdelrSZdIKsauVdF6er5jSkl9625gHvcUHRfGDDjaXPUM/MhgnbRAO6Y1bzWxFd
 DUELgVAztu4dkHHr4xu0HSfkduLQweqtnNAG2ZaHIG0ceDeEi1AqQsLHVca623ql
 dc5AD7XLGU3r0s8C88aVPE1zNOvMgs9OtD6EaVJHTg7YI6PC7R/DRSWEuxid3ana
 xKhGMzBFuE4cnIgYuhTaDInA4DoxrJtG7YxZURmKWuFw1Mgh+QS9VNY3d5n3QDM=
 =zXMQ
 -----END PGP SIGNATURE-----
 

Fix ClickEventCount for parallelism higher than 1

Following the [Operations Tutorial](https://nightlies.apache.org/flink/flink-docs-release-1.17/docs/try-flink/flink-operations-playground/#upgrading--rescaling-a-job) you're asked to start the job with parallelism 3 but I wasn't getting any output when I did that.

The reason I realized is [idle partitions](https://nightlies.apache.org/flink/flink-docs-master/docs/dev/datastream/event-time/generating_watermarks/#dealing-with-idle-sources) on the Kafka input because all input goes into partition 0.

By adding `withIdleness` to the WatermarkStrategy we are ensuring the job can work with higher parallelism and idle partitions.

Bonus: this also fixes the recovery in the [normal case](https://nightlies.apache.org/flink/flink-docs-release-1.17/docs/try-flink/flink-operations-playground/#step-2-introducing-a-fault). Without it I was seeing outputs of less than 1k clicks per window when resuming the job, whereas now it's the expected 1k every time.