Architecture and method for executing program builds

• US 7,676,788 B1
• Filed: 03/25/2003
• Issued: 03/09/2010
• Est. Priority Date: 03/25/2003
• Status: Active Grant

• Current Assignee: Cloudebees Incorporated
• Original Assignee: Electric Cloud, Inc.
• Inventors: Graham-Cumming, John, Delmas, Sven, Ousterhout, John, Muzaffar, Usman, Stanton, Scott, Melski, John Eric
• Primary Examiner(s): Zhen; Wei Y
• Assistant Examiner(s): Nguyen; Phillip H
• Application Number:
  US10/397,995
• Time in Patent Office: 2,541 Days
• Field of Search: 717/106
• US Class Current: 717/106
• CPC Class Codes: G06F 8/71 Version control security arrangements therefor G06F21/57; Configuration management
• CPC Class Codes: G06F 9/44521 Dynamic linking or loading; Link editing at or after load time, e.g. Java class loading
• CPC Class Codes: G06F 9/44552 Conflict resolution, i.e. enabling coexistence of conflicting executables

Abstract

A method is described comprising: scheduling jobs for a program build to execute in parallel across a plurality of nodes; predicting the files required to complete each of the jobs; and preloading the files to each node prior to the execution of each job.

Claims

1. A system comprising:

a plurality of build machines, wherein each build machine includes a processor to execute instructions and a memory coupled with the processor to store instructions that when executed cause the processor to schedule jobs for a program build invoked by that build machine;

a computing device including a memory to store program code and a processor to execute the program code to generate a cluster manager to dynamically allocate different ones of a plurality of nodes from a node cluster to different ones of the plurality of build machines for the program builds based on a specified configurable node allocation policy, wherein for each build machine the node allocation policy indicates which one or more of the nodes from the node cluster to allocate to that build machine based on a type of the program build invoked by that build machine;

the plurality of nodes to execute the jobs for each of the program builds in parallel according to the schedule generated by the build machines; and

a job scheduler module to analyze file dependency information contained within a first file, determine, based on the analysis, which jobs may be executed in parallel, and schedule the jobs which may be executed in parallel to execute in parallel across the plurality of nodes allocated by the cluster manager; and

a usage tracking module to collect file usage information from each of the jobs, the nodes providing the file usage information to the build machine following the completion of each of the jobs, the usage tracking module to supplement the file dependency information with the file usage information following execution of one or more of the jobs;

wherein the job scheduler module analyzes both the file dependency information and the file usage information contained in the first file and schedules jobs to execute in parallel for a subsequent program build based on the analysis of the file dependency information and the file usage information.

2. The system as in claim 1 wherein the first file is a Makefile.

3. The system as in claim 1 wherein the build machine further comprises:

a conflict detection module to detect a conflict in the program build by comparing versions of files used to complete a job, as indicated by the file usage information, with the versions of files that should have been used to complete the job.

4. The system as in claim 3 wherein the build machine further comprises:

a conflict resolution module to discard the results of the job that created the conflict and re-execute the job on one of the nodes using the correct file versions.

5. The system as in claim 4 wherein the correct file versions are identified as the versions used in a non-parallel program build.

6. The system as in claim 1 wherein the cluster manager comprises:

a node status module to monitor the status of each of the nodes, wherein the cluster manager allocates only those nodes which are operating property.

7. The system as in claim 1 wherein the cluster manager comprises a management interface to allow system administrators to set the node allocation policy.

8. The system as in claim 1 wherein the node allocation policy comprises:

allocating nodes to build machines requesting nodes during certain specified periods of time; and

reserving nodes for only certain build machines during other specified periods of time.

9. The system as in claim 8 wherein the other specified periods of time are periods of time during which clean builds are executed on the system.

21. The system as in claim 1, wherein the node allocation policy indicates the cluster manager to allocate all nodes from the node cluster to a build machine invoking a nightly build, and wherein the node allocation policy indicates the cluster manager to allocate a portion of the nodes from the node cluster to multiple build machines if each build machine is invoking an individual build.

10. A computing device including a memory to store program code and a processor to execute the program code to generate a cluster manager to be used within a system to perform program builds, the cluster manager comprising:

a node status module to monitor the status of each of a plurality of processing nodes of a node cluster; and

a node allocation module to dynamically allocate the nodes, or a subset thereof, to one or more build machines upon request from those build machines, the allocation of nodes based on a specified configurable node allocation policy, wherein for each requesting build machine, the node allocation policy indicates which nodes from the node cluster to allocate to that build machine based on the number of requesting build machines, wherein in response to the nodes being allocated to the one or more build machines, the nodes to execute jobs provided by the one or more build machines and to store results associated with the executed jobs; and

a job scheduler module to analyze file dependency information contained within a first file, determine, based on the analysis, which jobs may be executed in parallel, and schedule the jobs which may be executed in parallel to execute in parallel across the plurality of nodes; and

a usage tracking module to collect file usage information from each of the jobs, the nodes providing the file usage information to the build machine following the completion of each of the jobs, the usage tracking module to supplement the file dependency information with the file usage information following execution of one or more of the jobs;

wherein the job scheduler module analyzes both the file dependency information and the file usage information contained in the first file and schedules jobs to execute in parallel for a subsequent program build based on the analysis of the file dependency information and the file usage information.

11. The cluster manager as in claim 10 further comprising:

a management interface to allow system administrators to set the node allocation policy.

12. The cluster manager as in claim 11 wherein the node allocation policy comprises:

allocating nodes to build machines requesting nodes during certain specified periods of time; and

reserving nodes for only certain build machines during other specified periods of time.

13. The system as in claim 12 wherein the other specified periods of time are periods of time during which clean builds are executed on the system.

14. The system as in claim 10 wherein the node allocation module allocates nodes to build machines based on the operational status of each of the nodes.

19. The cluster manager of claim 10, wherein the results are to be stored on a mass storage device.

15. A build machine comprising:

a processor to execute instructions;

a memory coupled with the processor, the memory having instructions that when executed cause the processor to perform operations to generate the following;

a central build module to request a plurality of nodes to be used during a program build, the nodes being dynamically selected from a node cluster and allocated to the build machine based on a specified configurable node allocation policy, wherein the node allocation policy indicates which one or more of the nodes from the node cluster to allocate to the build machine based on a type of the program build invoked by the build machine;

one or more worker modules comprised of a job scheduler module to schedule and coordinate the execution of jobs for the program build across the plurality of nodes; and

a terminator module to determine whether the results of each job are valid and to correct out-of-order execution of jobs prior to writing the job results to a database; and

the job scheduler module to analyze file dependency information contained within a first file, determine, based on the analysis, which jobs may be executed in parallel, and schedule the jobs which may be executed in parallel to execute in parallel across the plurality of nodes; and

a usage tracking module to collect file usage information from each of the jobs, the nodes providing the file usage information to the build machine following the completion of each of the jobs, the usage tracking module to supplement the file dependency information with the file usage information following execution of one or more of the jobs;

wherein the job scheduler module analyzes both the file dependency information and the file usage information contained in the first file and schedules jobs to execute in parallel for a subsequent program build based on the analysis of the file dependency information and the file usage information.

16. The build machine as in claim 15 wherein the worker modules comprise:

a parser module to parse dependency information from the first file and generate a set of construction rules;

a job scheduler to schedule the jobs across the plurality of nodes based on the construction rules; and

a plurality of executor modules to coordinate the execution of each of the scheduled jobs at each of the nodes.

17. The build machine as in claim 16 wherein the first file containing the dependency information comprises a Makefile.

18. The build machine as in claim 16 wherein the first file containing the dependency information comprises one or more recursive Makefiles extracted from one or more of the jobs.

20. The build machine as in claim 15, further comprising a mass storage device to store the database.

22. The build machine of claim 15, wherein the terminator module is to determine whether the results of each job are valid by comparing the versions of files used to complete the job with the versions of files that should have been used to complete the job, and wherein the terminator module is to correct out-of-order execution of jobs by identifying the file versions that should have been used to complete the job and re-executing the job with those file versions.