salmer / loganalyser

The application for the analysis of game server logs.

LogAnalyser

This is an application that I implemented as a test task during the interviewing process in 2017. I rewrote it in 2021 to use it as a demo repository.

Table of contents

• Problem statement
• Input data description
• Application requirements
• Output requirements
• Project structure
• Execution manual
• Exception handling
• Performance metrics

Problem statement

The game server produces logs in JSON format. The main goal is to parse the collection of logs and aggregate data in the requested style to the reporting file.

Input data description

Any log file contains a lot of lines in JSON format. The JSON format describes user actions:

{
	"ts_fact" : 1489494303,
	"fact_name": "fact1",
	"actor_id": 111222,
   
	"props":
	{
   	"prop1": 11,
   	"prop2": 22,
	…
   	"prop10": 1010,
	}
}

Description:

Parameter	Description
ts_fact	The timestamp of an action
fact_name	The name of an action
actor_id	The user Id that produced an action
props	Properties of an action: prop1...prop10

Application requirements

• It must be running as a console application;
• It must be possible to analyze log files in multithreading mode;
• It must be possible to configure execution parameters via the command line:
  • The logs directory;
  • The number of files in the directory;
  • The number of threads running at the same time;
• It must be written in C++ (avoid C-style);
• The key points is performance.

Output requirements

The application must produce an output file: arg.log. It should contain the following data:

• For each UTC+0:
  • For each fact_name:
    • For each set of props:
      • The total amount of events found in logs.

Project structure

Stack

• C++17;
• MSVS 2019 (v142);
• fmtlib/fmt;
• tencent/rapidjson.

Solution ogranization

The solution contains 3 projects:

• LogAnalyserLib - the library contains a collection of functional units:
  • CommandLineParser - parses command-line arguments set by a user;
  • ConsoleAppLogger - writes messages to the console. The wrapped is based on the fmtlib/fmt;
  • DataMerger - merges and calculated aggregated JSON data from all input files in an asynchronous way;
  • LogReader - reads each log file separately. It can be used in an asynchronous way. The JSON is parsed by means of tencent/rapidjson;
  • PlayerActionInfo - contains a collection of application data structures;
  • PlayerInfoAnalyser - analyses raw data grabbed from a single file;
  • ResultWriter - produces the output file and writes data to it.
• LogAnalyserTests - contains a collection of unit-tests to test the funcional uints of the LogAnalyserLib;
• LogAnalyserBin - contains a "glue-code" that implements the applictaion algorithm;
• testdata - contains test samples for the application.

Execution manual

The application can be executed via the command line:

> LogAnalyserBin.exe -d ..\\testFolder -n 10 -t 4

Description:

Parameter	Description
-d	The directory with a collection of log files
-n	The number of files in the directory
-t	The number of active threads

Exception handling

• The application based on the exception mechanism. In case of any issues, an exception will be thrown and the application will be aborted. The description of an exception can be found in the command line window.
• The application returns 0 in case of a successful run otherwise it returns -1.

Performance metrics

The application was tested with the next hardware set:

• Core: Intel Core i5 9300H 2.40GHz;
• RAM: 16Gb;
• OS: Microsoft Windows 10 20H2.

The single file description:

• Size: ~760Mb
• Lines: > 4 millions.

The data set contains 4 files. The test results can be found below:

JSON parser	Single file(1 thread)	All files (4threads)
nlohmann/json	~35sec.	~51sec.
tencent/rapidjson	~14sec.	~25sec.