Author: Qiu from Xian (Chi Hill)
In Flink, the state guarantee reliability supported by Checkpoint, in the case where the job appears failover, Flink will recover from the recent success of Checkpoint. In reality, we may experience failures Checkpoint, Checkpoint slow or situation, we will be unified chat Flink in unusual circumstances Checkpoint (including failures and slow), and possible causes and troubleshooting ideas.
1. Checkpoint Process Description
First, we need to understand the entire process Flink in Checkpoint is how, in the understanding of the whole process, we can at the time of the problem, better positioning analysis.
From the graph we can know, Flink's Checkpoint includes the following sections:
- JM trigger checkpoint
- Source received trigger checkpoint