Event panel & insights
The Unravel intelligence (UI) engine helps you manage your applications more efficiently by providing insights into their run. The UI engine gives its insights and tuning suggesting via the Events Panel. Not all UI engine insights results in concrete recommendations, so to take full advantage you must read the efficiency panel. There is not a one-to-one correspondence between the event and recommendation number. A single event might lead to none or many recommendations.
Recommendations
Lists the parameters to change, shows their current and recommended value.
Efficiency
The efficiency list details the inefficiencies. The UI engine might:
Make a recommendation and note the expected result from such a change.
Make a configuration suggestion.
Suggest where to look to increase efficiency.
Below are two examples. Each type of job and instance of a job has events relevant to that particular job and instance.
MapReduce job example
This MapReduce job is part of a Hive Query. In this example the UI engine lists list four events and has three recommendations.
Recommendations
Efficiency 1: Used too many reducers
Resulted in the one recommendation (#1).
Efficiency 2: Reduce tasks that start before map phase finishes.
Resulted in one suggestion.
Efficiency 3: Too many mappers.
Resulted in the two recommendations (#2 and #3).
Efficiency 4: Large data shuffle from map to reduce.
Resulted in a suggestion.
Tez DAG example
This Tez DAG job is part of a Hive Query. In this example the UI engine lists list three events and has four recommendations.
Recommendations
Efficiency 1: Tez DAG map vertex used too many tasks.
Resulted in two suggestions (#3 and #4) and explanation of the problem.
Efficiency 2: Tez DAG reducer vertex used too many tasks.
Resulted in one recommendation (#1).
Efficiency 3: hive.exec.parallel is set to false.
Resulted in one recommendation (#2).