The fundamental purpose of crash groups is to isolate the crash of a specific element from the rest of the pipeline. If an element is likely to crash (e.g., it interacts with unstable external resources), it is usually assigned to a crash group containing all elements inextricably linked to its operation. This prevents a localized failure from bringing down the entire system and allows for the controlled recovery of specific logical units. This approach also ensures that components generally needing a restart or termination are cleaned up correctly.
0 commit comments