A grossly oversimplified version is as follows:

You take genomic DNA as input. Some of this DNA is cleaned up, sheared, size selected, has adaptors added, and is often amplified, which results in a library. This library DNA is then sequenced. If you still don't have adequate coverage of your genome, you may wish to prepare another library, starting again from the genomic DNA.

Since these libraries are from the same sample, the reads are often just concatenated together into one big bam. The readgroup and library tags let you trace each read back to it's origin.