Hi, In our project we recently upgraded from Flink 1.9 to 1.20.
After the upgrade we saw a problem of two subtasks (in a task that reads several Kafka topics) out of four were not producing any records. We are reading three different topics each having two partitions. All of the partitions are producing messages continuously. This was fixed by changing slot count from four to two. I am glad to have this fixed but I was left wondering why. Moreover, what would be a reasonable "algorithm" of setting the slot count in general? Eg. "set slot count to partition count of the topic that has the most partitions". PS. I requested access to Flink Slack. If you, the reader, can help me with this, I would appreciate it. Thanks and regards, Lauri Mäkinen
