Why does shrinking the transaction log generate lots of log (LOP_SHRINK_NOOP) and not shrink the log?

I just tried shrinking an overly big log file (FULL recovery model). This operation is not expected to always succeed and indeed the log was not shrunk.

But I observed that a lot of new log was generated. At a log file size of about 20GB each shrink attempt generated a few hundred MB of new log.

Reading out the log I found that all of that space was for LOP_SHRINK_NOOP log records. The web does not have information about this log record type.

DBCC LOGINFO shows that there is a mix of active and unused VLFs. There is one used VLF at the end of the file ( Status = 2).

Why does attempting to shink the log sometimes...

1. ...cause lots of new log data?


2. ...not shrink the log file although apparently work was performed on it?


3. ...generate LOP_SHRINK_NOOP records? What does that log record type mean and when is it expected to appear?