Methods and systems of function-specific tracing

• US 9,658,941 B2
• Filed: 01/28/2013
• Issued: 05/23/2017
• Est. Priority Date: 01/28/2013
• Status: Active Grant

• Current Assignee: Rackspace US Incorporated (Apollo Global Management, Inc.)
• Original Assignee: Rackspace US Incorporated (Apollo Global Management, Inc.)
• Inventors: Dietz, Matthew Charles, Voccio, Paul
• Primary Examiner(s): Zhen, Li B
• Assistant Examiner(s): Soltanzadeh, Amir
• Application Number:
  US13/752,234
• Publication Number:
  US 20140215444A1
• Time in Patent Office: 1,576 Days
• CPC Class Codes: G06F 11/3636 by tracing the execution of the program
• CPC Class Codes: G06F 11/3644 by instrumenting at runtime

Abstract

A system and methods are provided for function-specific tracing of a program. In one embodiment, a method includes generating a trace profile identifying one or more functions of a target program, wherein the trace profile identifies one or more functions to trace and depth of tracing for each function to be traced, loading the trace profile and the target program, identifying traced functions in the target program based on the trace profile, patching the target program to call a trace parameter for one or more functions, wherein traced functions are declared at runtime, and observing function calls for traced functions of the application. In this regard, individual functions are traced and debugged on a function-by-function basis without modifying the code or pre-arranging functions so they are traceable. As such, the scope of tracing may be dynamically limited to yield only information that is desired.

Claims

1. A method of function-specific tracing in a distributed application, the method comprising:

generating a trace profile identifying one or more functions of a target program of the distributed application, wherein the trace profile identifies one or more functions to trace using a binary-relative address for each function to be traced and depth of tracing for each function to be traced;

patching, by a tracer process, the target program of the distributed application in accordance with the generated trace profile to trace the one or more functions to the identified depth for each of the one or more functions;

loading, into a compute controller in a distributed computing system, the patched target program;

executing, by the compute controller, the patched target program;

identifying, by the compute controller, traced functions in the patched target program based on a binary-relative address, wherein the traced functions are identified at runtime;

observing function calls for traced functions of the target program, wherein observing function calls for traced functions of the target program includes executing the patched target program by a first process of the compute controller, and wherein observing function calls for traced functions of the patched target program is performed by a second process;

constructing a per-process call stack for each traced function;

combining, by the compute controller, the per-process call stacks to generate a distributed call stack, wherein the distributed call stack is utilized to show call flows among several functions within and between processes, permitting tracing of calls from receiving of requests to the distributed application; and

fixing, by the compute controller, a call stack after execution of the traced functions.

2. The method of claim 1, wherein the trace profile is a configuration file or passed as part of calling the tracer.

3. The method of claim 1, wherein the target program is an unmodified version of code for the program.

4. The method of claim 1, wherein traced functions are identified based on a description for each function to be traced in the trace profile.

5. The method of claim 1, wherein a traced function is declared via its path, and identification of a traced function is based on a declared path.

6. The method of claim 1, wherein patching includes patching the target program to call into a tracer routine at the entrance and exit of each function.

7. The method of claim 1, wherein patching includes wrapping a traced function with a decorator, and wherein observing includes observing a decorated function.

8. The method of claim 1, wherein observing a traced function includes tracing bytecode before execution, and wherein function calls are inserted into the target program based on bytecode manipulation.

9. The method of claim 1, wherein observing includes identifying traced functions by one or more of a trace profile by way of output files for debugging, tracing a symbolic call stack, and launching a rebase to call into a trace library.

10. The method of claim 1, wherein observing includes recording arguments passed by traced functions based on the depth of tracing for each traced function.

11. The method of claim 1, wherein observing includes recording entrance and exit values of observed functions.

12. The method of claim 1, further comprising outputting one or more of a list of the function calls for each traced function and call stack of traced function calls.

13. A system comprising:

one or more processors;

one or more memory devices; and

plurality of modules including;

a trace profile module to generate a trace profile identifying one or more functions of a target program, wherein the trace profile identifies one or more functions to trace using a binary-relative address for each function to be traced and depth of tracing for each function to be traced;

a tracer module to patch the target program in accordance with the generated trace profile to trace the one or more functions to the identified depth for each of the one or more functions;

a module to load, into a controller module of a distributed computing system, the patched target program; and

a controller module to;

execute the patched target program;

identify traced functions in the patched target program based on the trace profile and based on a binary-relative address, wherein the traced functions are declared at runtime;

observe function calls for traced functions of the target program by executing the patched target program by a first process of the controller module;

cause observing of function calls for traced functions of the patched target program;

construct a per-process call stack for each traced function;

combine the per-process call stacks to generate a distributed call stack, wherein the distributed call stack is utilized to show call flows among several functions within and between processes, permitting tracing of calls from receiving of requests to the target program; and

fix a call stack after execution of the traced functions.

14. The system of claim 13, wherein the trace profile is a configuration file or passed as part of calling the tracer.

15. The system of claim 13, wherein the target program is an unmodified version of code for the program.

16. The system of claim 13, wherein traced functions are identified based on a description for each function to be traced in the trace profile.

17. The system of claim 13, wherein a traced function is declared via its path, and identification of a traced function is based on a declared path.

18. The system of claim 13, wherein patching includes patching the target program to call into a tracer routine at the entrance and exit of each function.

19. The system of claim 13, wherein patching includes wrapping a traced function with a decorator, and wherein observing includes observing a decorated function.

20. The system of claim 13, wherein observing a traced function includes tracing bytecode before execution, and wherein function calls are inserted into the target program based on bytecode manipulation.

21. The system of claim 13, wherein observing includes identifying traced functions by one or more of a trace profile by way of output files for debugging, tracing a symbolic call stack, and launching a rebase to call into a trace library.

22. The system of claim 13, wherein observing includes recording arguments passed by traced functions based on the depth of tracing for each traced function.

23. The system of claim 13, wherein observing includes recording entrance and exit values of observed functions.

24. The system of claim 13, wherein the controller module is further configured to output one or more of a list of the function calls for each traced function and call stack of traced function calls.

25. A non-transitory computer-accessible storage medium storing program instructions that when executed by a computer cause the computer to implement a function-specific tracing of a program, including instructions to:

generate a trace profile identifying one or more functions of a target program, wherein the trace profile identifies one or more functions to trace using a binary-relative address for each function to be traced and depth of tracing for each function to be traced;

patch, by a tracer process, the target program in accordance with the generated trace profile to trace the one or more functions to the identified depth for each of the one or more functions;

load, into a compute controller in a distributed computing system, the patched target program;

execute, by the compute controller, the patched target program;

identify, by the compute controller, traced functions in the patched target program based on a binary-relative address, wherein the traced functions are identified at runtime;

observe function calls for traced functions of the target program by executing the patched target program by a first process of the controller module and observing function calls for traced functions of the patched target program by a second process of the controller module;

construct a per-process call stack for each traced function;

combine, by the compute controller, the per-process call stacks to generate a distributed call stack, wherein the distributed call stack is utilized to show call flows among several functions within and between processes, permitting tracing of calls from receiving of requests to the target program; and

fix, by the compute controller, a call stack after execution of the traced functions.