CSE444 Lab 5: Rollback and Recovery

Assigned: Friday, May 18, 2012
Due: Friday, June 1, 2012

0. Introduction

In this lab you will implement log-based rollback for aborts and log-based crash recovery. We supply you with the code that defines the log format and appends records to a log file at appropriate times during transactions. You will implement rollback and recovery using the contents of the log file.

The logging code we provide generates records intended for physical whole-page undo and redo. When a page is first read in, our code remembers the original content of the page as a before-image. When a transaction updates a page, the corresponding log record contains that remembered before-image as well as the content of the page after modification as an after-image. You'll use the before-image to roll back during aborts and to undo loser transactions during recovery, and the after-image to redo winners during recovery.

We are able to get away with doing whole-page physical UNDO (while ARIES must do logical UNDO) because we are doing page level locking and because we have no indices which may have a different structure at UNDO time than when the log was initially written. The reason page-level locking simplifies things is that if a transaction modified a page, it must have had an exclusive lock on it, which means no other transaction was concurrently modifying it, so we can UNDO changes to it by just overwriting the whole page.

Your BufferPool already implements abort by deleting dirty pages, and pretends to implement atomic commit by forcing dirty pages to disk only at commit time. Logging allows more flexible buffer management (STEAL and NO-FORCE), and our test code calls BufferPool.flushAllPages() at certain points in order to exercise that flexibility.

1. Getting started

You should begin with the code you submitted for Lab 3 (if you did not submit code for Lab 3, or your solution didn't work properly, contact us to discuss options.)

You'll need to modify some of your existing source and add a few new files. Here's what to do:

Make a backup of your Lab 3 solution by typing :

$ tar -cvzf CSE444-lab3-submitted.tar.gz CSE444-lab3

Rename your lab directory to CSE444-lab5
```
$ mv CSE444-lab3 CSE444-lab5
```
Change to the directory that contains your top-level simpledb code:
```
$ cd CSE444-lab5
```
Download the new tests and skeleton code for Lab 5 from http://www.cs.washington.edu/education/courses/cse444/12sp/labs/lab5/CSE444-lab5-supplement.tar.gz:
```
$ wget
http://www.cs.washington.edu/education/courses/cse444/12sp/labs/lab5/CSE444-lab5-supplement.tar.gz
```
Extract the new files for Lab 5 by typing:
```
tar -xvzf CSE444-lab5-supplement.tar.gz
```
This will not overwrite any existing files, but will just add a new test file to the test/simpledb/systemtest directory.
Now make the following changes to your existing code:
1. Insert the following lines into BufferPool.flushPage() before your call to writePage(p), where p is a reference to the page being written:
```
        // append an update record to the log, with 
        // a before-image and after-image.
        TransactionId dirtier = p.isDirty();
        if (dirtier != null){
          Database.getLogFile().logWrite(dirtier, p.getBeforeImage(), p);
          Database.getLogFile().force();
        }
```
  This causes the logging system to write an update to the log. We force the log to ensure the log record is on disk before the page is written to disk.
2. Your BufferPool.transactionComplete() calls flushPage() for each page that a committed transaction dirtied. For each such page, add a call to p.setBeforeImage() after you have flushed the page:
```
        // use current page contents as the before-image
        // for the next transaction that modifies this page.
        p.setBeforeImage();
```
  After an update is committed, a page's before-image needs to be updated so that later transactions that abort rollback to this committed version of the page. (Note: We can't just call setBeforeImage() in flushPage(), since flushPage() might be called even if a transaction isn't committing. Our test case actually does that! If you implemented transactionComplete() by calling flushPages() instead, you may need to pass an additional argument to flushPages() to tell it whether the flush is being done for a committing transaction or not. However, we strongly suggest in this case you simply rewrite transactionComplete() to use flushPage().)
After you have made these changes, do a clean build (ant clean; ant from the command line, or a "Clean" from the "Project" menu in Eclipse.)

At this point your code should pass the first three sub-tests of the LogTest systemtest, and fail the rest:

% ant runsystest -Dtest=LogTest
    ...
    [junit] Running simpledb.systemtest.LogTest
    [junit] Testsuite: simpledb.systemtest.LogTest
    [junit] Tests run: 10, Failures: 0, Errors: 7, Time elapsed: 0.42 sec
    [junit] Tests run: 10, Failures: 0, Errors: 7, Time elapsed: 0.42 sec
    [junit] 
    [junit] Testcase: PatchTest took 0.057 sec
    [junit] Testcase: TestFlushAll took 0.022 sec
    [junit] Testcase: TestCommitCrash took 0.018 sec
    [junit] Testcase: TestAbort took 0.03 sec
    [junit]     Caused an ERROR
    [junit] LogTest: tuple present but shouldn't be
    ...

If you don't see the above output from ant runsystest -Dtest=LogTest, something has gone wrong with the un-tarring of CSE444-lab5-supplement.tar.gz or the changes you made are somehow incompatible with your existing code. You should figure out and fix the problem before proceeding; ask us for help if necessary.

2. Rollback

Read the comments in LogFile.java for a description of the log file format. You should see in LogFile.java a set of functions, such as logCommit(), that generate each kind of log record and append it to the log.

Your first job is to implement the rollback() function in LogFile.java. This function is called when a transaction aborts, before the transaction releases its locks. Its job is to un-do any changes the transaction may have made to the database.

Your rollback() should read the log file, find all update records associated with the aborting transaction, extract the before-image from each, and write the before-image to the table file. Use raf.seek() to move around in the log file, and raf.readInt() etc. to examine it. Use readPageData() to read each of the before- and after-images. You can use the map tidToFirstLogRecord (which maps from a transaction id to an offset in the heap file) to determine where to start reading the log file for a particular transaction. You will need to make sure that you discard any page from the buffer pool whose before-image you write back to the table file.

As you develop your code, you may find the Logfile.print() method useful for displaying the current contents of the log.

Exercise 1: LogFile.rollback()

Implement LogFile.rollback().

After completing this exercise, you should be able to pass the TestAbort and TestAbortCommitInterleaved sub-tests of the LogTest system test.

3. Recovery

If the database crashes and then reboots, LogFile.recover() will be called before any new transactions start. Your implementation should:

Read the last checkpoint, if any.
Scan forward from the checkpoint (or start of log file, if no checkpoint) to build the set of loser transactions. Re-do updates during this pass. You can safely start re-do at the checkpoint because LogFile.logCheckpoint() flushes all dirty buffers to disk.
Un-do the updates of loser transactions.

Exercise 2: LogFile.recover()

Implement LogFile.recover().

After completing this exercise, you should be able to pass all of the LogTest system test.

4. Logistics

You must submit your code (see below) as well as a short (1 page, maximum) writeup describing your approach. This writeup should:

Describe any design decisions you made, including anything that was difficult or unexpected.
Discuss and justify any changes you made outside of LogFile.java.

4.1. Collaboration

All CSE 444 labs are to be completed INDIVIDUALLY! However, you may discuss your high-level approach to solving each lab with other students in the class.

4.2. Submitting your assignment

To submit your code, please create a CSE444-lab5.tar.gz tarball (such that, untarred, it creates a CSE444-lab5 directory with your source code at CSE444-lab5/src/java/simpledb). Include your individual writeup in the tarball that you submit:

$ cp lab5_writeup.pdf CSE444-lab5

$ tar -cvzf CSE444-lab5.tar.gz CSE444-lab5

Please do not use the ant handin target to create your submission archive.

Submit your tarball for the Lab 5 assigment to the dropbox.. You may submit your code multiple times; we will use the latest version you submit that arrives before the deadline (before 11:59pm on the due date). Please submit your writeup as a PDF, plain text file, or word document (.doc or .docx).

4.3. Submitting a bug

SimpleDB is a relatively complex piece of code. It is very possible you are going to find bugs, inconsistencies, and bad, outdated, or incorrect documentation, etc.

We ask you, therefore, to do this lab with an adventurous mindset. Don't get mad if something is not clear, or even wrong; rather, try to figure it out yourself or send us a friendly email. Please submit (friendly!) bug reports to the course staff. When you do, please try to include:

A description of the bug.
A .java file we can drop in the src/simpledb/test directory, compile, and run.
A .txt file with the data that reproduces the bug. We should be able to convert it to a .dat file using PageEncoder.

You can also post on the class message board if you feel you have run into a bug.

4.4 Grading

50% of your grade will be based on whether or not your code passes the system test suite we will run over it. These tests will be a superset of the tests we have provided; the tests we provide are to help guide your implementation, but they do not define correctness. Before handing in your code, you should make sure produces no errors (passes all of the tests) from both ant test and ant systemtest.

Important: Before testing, we will replace your build.xml, HeapFileEncoder.java, and the entire contents of the test/ directory with our version of these files. This means you cannot change the format of the .dat files! You should also be careful when changing APIs and make sure that any changes you make are backwards compatible. In other words, we will untar your tarball, replace the files mentioned above, compile it, and then grade it. It will look roughly like this:

$ gunzip CSE444-lab5.tar.gz
$ tar xvf CSE444-lab5.tar
$ cd CSE444-lab5
[replace build.xml, HeapFileEncoder.java and test]
$ ant test
$ ant systemtest
[additional tests]

If any of these commands fail, we'll be unhappy, and, therefore, so will your grade.

An additional 50% of your grade will be based on the quality of your writeup and our subjective evaluation of your code.

We've had a lot of fun designing this assignment, and we hope you enjoy hacking on it!

Last modified: Wed May 2 22:30:00 EDT 2012