Page History

...

3) Suppose that the first consumer crashed yet again, in this case, another lag could form as a result of this failure. We will repeat the same set of actions described previously, but without creating a new thread. In this case, we would have the first consumer again seek to the end of the log. Meanwhile, the second consumer will start processing the offset lag that results accordingly.

Image Added

With this design, you could summarize the two consumer threads as follows:

...