DEBUGGING IN A CLUSTER PROCESSING NETWORK

• US 20120124426A1
• Filed: 11/12/2010
• Published: 05/17/2012
• Est. Priority Date: 11/12/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Maybee, Paul, Moth, Daniel, Panapakkam, Anitha
• Granted Patent: US 8,412,984 B2
• Time in Patent Office: Days
• US Class Current: 714/38.11
• CPC Class Codes: G06F 11/0715 in a system implementing multitasking multitasking per se G06F9/46
• CPC Class Codes: G06F 11/0757 by exceeding a time limit, i.e. time-out, e.g. watchdogs
• CPC Class Codes: G06F 11/362 Software debugging

Abstract

A technology is described for debugging in a cluster processing network. A scheduler can dispatch a process that is part of the cluster job for execution. Further, a compute node can be used to execute the process dispatched by the scheduler to the compute node. A debugger can be activated in response to an unhandled suspension event in the process on the compute node. In addition, the debugger can send notification messages regarding the unhandled suspension event. A job monitor can receive a notification from the debugger that an unhandled suspension event has occurred. The notification can be displayed to a user via the job monitor.

Claims

1. A system for debugging in a cluster processing network, comprising:

a scheduler to dispatch a process that is part of the cluster job for execution;

a compute node to execute the process dispatched by the scheduler to the compute node;

a debugger activated in response to an unhandled suspension event for the process on the compute node, wherein the debugger sends a notification message regarding the unhandled suspension event; and

a job monitor to receive the notification message from the debugger that an unhandled suspension event has occurred.

2. The system as in claim 1, wherein the debugger informs the job monitor that the process that is part of a cluster job is waiting for debugging to begin.

3. The system as in claim 1, further comprising a user interface in the job monitor to display a notification that the process on the compute node is waiting for an attachment of a debugger.

4. The system as in claim 1, further comprising a user interface in the job monitor to display a notification of an activation of the debugger on the compute node.

5. The system as in claim 1, wherein the debugger accesses environmental variables in order to determine an action to take when the unhandled suspension event is received.

6. The system as in claim 5, wherein environmental variables determine the debugger'"'"'s actions when the unhandled suspension event is received and the environmental variables are selected from the group consisting of performing a debug file dump, waiting for an attach for a crashed process, and waiting for an attach for a process that has a debug break.

7. The system as in claim 5, wherein the environmental variables determine whether a debugger times out waiting for a debugger attach after a pre-defined time limit.

8. The system as in claim 5, wherein the debugger writes a crash dump for the process to a pre-defined location associated with the compute node.

9. The system as in claim 1, wherein the debugger opens a handle to the process and waits for the process to be debugged.

10. The system as in claim 1, further comprising a head node to manage a cluster job processed on the cluster processing network, wherein the head node contains the job monitor.

11. A method for debugging of a process in a cluster job in a cluster processing network, comprising:

dispatching a process to execute on a compute node using a scheduler;

executing the process that is part of the cluster job on the compute node;

activating a debugger on the compute node in response to an unhandled suspension event for the process; and

sending a notification from the debugger to a job monitor that an unhandled suspension event has occurred.

12. The method as in claim 11, further comprising informing the job monitor that a process for the cluster job is waiting for debugging to begin.

13. The method as in claim 11, further comprising displaying a notification about an activation of the debugger using a user interface in the job monitor.

14. The system as in claim 11, further comprising accessing environmental variables with the debugger in order to determine an action to take upon receiving an unhandled exception.

15. The system as in claim 11, further comprising opening a handle to the process and waits for the process to be debugged.

16. The method as in claim 11, further comprising starting the process on the compute node with just-in-time debugging configuration flags.

17. A method for just-in-time debugging in a cluster processing network, comprising:

starting a process on a compute node with just-in-time debugging configuration flags;

informing an operating system on the compute node that an unhandled exception occurred for a process executing on the compute node;

starting a just-in-time debugger based on a configuration setting read by the operating system, wherein the just-in-time debugger attaches to the process;

reading the just-in-time debugging configuration flags to determine an action for the just-in-time debugger to take; and

sending a message to a job manager that a process which is part of a cluster job is waiting for debugging to begin.

18. The method as in claim 17, further comprising taking a crash dump of the process based on the debugging configuration flags.

19. The method as in claim 17, further comprising opening a handle to the process and waiting for the process to be debugged based on the debugging configuration flags.

20. The method as in claim 17, wherein informing a job manager on a head node that a process is waiting to be debugged further comprises informing a user that a process is waiting for debugging to begin via an email sent to a user.