By Alex Holmes
MapReduce patterns help you write effective code and make efficient use of your data and your Hadoop cluster. It can be just as useful to learn from anti-patterns, which are patterns that are commonly used but are either ineffective or worse, detrimental in practice. In this article based on chapter 13 of Hadoop in Practice, you can learn and laugh at mistakes that the author made in MapReduce on production clusters, which range from loading too much data into memory in tasks to going crazy with counters and bringing down the JobTracker.
MapReduce Anti-patterns (PDF)