victormejia wrote:

Just to make sure I am understanding this, Stage 1 is already implemented in the doRecover() method of the RecoveryMgr right? All we need to do is implement Stage 3, in which we have to iterate forward?

Yes, if you mean Stage 2.

Also, what do you mean by keeping one block of the log records in memory at any time?

When you iterate through the log file, you may only pin one block in memory at a time. You may keep additional data in memory to help with the forward iteration, but the size of the additional data must be small - it must be smaller than one block at least.