Usually due to insufficient resources, it is lower than the low water level. You can refer to the following inspection ideas:
1. When a single task cannot run out, it may be that the SQL task is large, resulting in insufficient resources. In this scenario, we can first analyze whether the SQL task can be split in large and small areas; if it includes large table calculations, analyze whether there is any partition design to better utilize the partition cutting capability.
2. Currently, query whether there are many tasks. If there are many concurrent executions, analyze whether the task can be arranged staggered.
3. To analyze the usage of BE memory, that is, whether the memory is released normally, whether there is a memory leak, resulting in resource shortage, you can conduct preliminary analysis in combination with memtrakcer