Global avoidance of hang states via priority inheritance in multi-node computing system

• US 8,429,657 B2
• Filed: 04/28/2008
• Issued: 04/23/2013
• Est. Priority Date: 04/28/2008
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Zoll, Michael, Pruscino, Angelo, Chan, Wilson
• Primary Examiner(s): PETRANEK, JACOB ANDREW
• Application Number:
  US12/150,393
• Publication Number:
  US 20090271794A1
• Time in Patent Office: 1,821 Days
• Field of Search: None
• US Class Current: 718/103
• CPC Class Codes: G06F 9/46 Multiprogramming arrangements
• CPC Class Codes: G06F 9/4818 Priority circuits therefor
• CPC Class Codes: G06F 9/5038 considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration scheduling strategies G06F9/4881 and subgroups
• CPC Class Codes: G06F 9/524 Deadlock detection or avoidance

Abstract

Systems, methods, and other embodiments associated with avoiding resource blockages and hang states are described. One example computer-implemented method for a clustered computing system includes determining that a first process is waiting for a resource and is in a blocked state. The resource that the first process is waiting for is identified. A blocking process that is holding the resource is then identified. A priority of the blocking process is compared with a priority the first process. If the priority of the blocking process is lower than the priority of the first process, the priority of the blocking process is increase. In this manner the blocking process can be scheduled for execution sooner and thus release the resource.

Claims

1. A computer-implemented method for a clustered computing system, the method comprising:

determining that a first process is waiting for a resource and that the first process is in a blocked state by accessing process state information from a shared global memory on a remote node, where the process state information identifies processes and a current state of each process, where the shared global memory is shared and remotely accessed by multiple nodes in the clustered computing system;

identifying the resource that the first process is waiting for;

identifying, from the shared global memory, a blocking process that is holding the resource and a processing state of the blocking process;

comparing a priority of the blocking process with a priority of the first process; and

selectively increasing the priority of the blocking process based, at least in part, on the priority and the process state information of the blocking process.

2. The computer-implemented method of claim 1 where if the first process is on a first node in the clustered computing system and the blocking process is on a second node, then sending a message to the second node to cause an operating system local to the second node to increase the priority of the blocking process.

3. The computer-implemented method of claim 1 where selectively increasing the priority of the blocking process includes increasing the priority if the priority of the blocking process is lower than the priority of the first process and the blocking process is not currently executing.

4. The computer-implemented method of claim 1 comprising:

selectively maintaining the blocking process in a currently executing state based, at least in part, on the processing state information of the blocking process, where selectively maintaining the blocking process in a currently executing state includes;

determining if the blocking process is currently executing based, at least in part, on the process state information for the blocking process; and

selectively instructing an operating system to prevent preemption of the blocking process by making the blocking process non-interruptable without increasing the priority of the blocking process or extending an execution time for the blocking process.

5. The computer-implemented method of claim 1 where the clustered computing system includes a plurality of computing nodes;

and

where identifying the blocking process includes identifying the blocking process by using the identity of the resource to search for the identity of the blocking process in the remote node from the process state information in the shared global memory on the remote node.

6. The computer-implemented method of claim 1, where determining that the first process is waiting includes a management process detecting a timeout by the first process and searching the process state information in the shared global memory for waiting processes that are blocked by another process.

7. The computer-implemented method of claim 1 further including:

determining whether a blocking chain is present where multiple processes are waiting for the resource, and if present, repeating the method for a subsequently blocked process.

8. The computer-implemented method of claim 1 where the method is initiated by the waiting process.

9. The computer-implemented method of claim 1, wherein selectively increasing the priority of the blocking process includes lending the priority of the first process to the blocking process and returning the lent priority when the blocking process releases the resource.

10. The computer-implemented method of claim 1, further comprising:

after the resource is released by the blocking process, passing the increased priority from the blocking process to a subsequent process that uses the resource, wherein the subsequent process is the first process.

11. A clustered computing system, comprising:

a plurality of computing nodes for executing processes, each of the computing nodes including a processor, an operating system, and a database instance that is configured as part of a clustered database system;

a shared global memory that is shared on a remote node of the computing nodes and accessed by the plurality of computing nodes in the clustered database system;

a blocking avoidance logic embodied in one or more of the plurality of nodes and when executed is configured to;

determine when a process in a first node of the computing nodes is in a wait state that is waiting for a resource, the process being a waiting process;

determine an identity and a processing state of a blocking process from the shared global memory on the remote node, where the blocking process is part of a second node of the computing nodes, to which the resource is allocated; and

selectively causing an operating system local to the blocking process to increase an execution priority of the blocking process based, at least in part, on an execution priority of the waiting process and the process state information of the blocking process.

12. The computing system of claim 11 where the shared global memory having stored therein:

process state information that identifies a current state and identifiers of processes on the clustered computing system; and

resource state information that identifies resources on the clustered computing system and identifies which process holds a resource.

13. The clustered computing system of claim 11 further including a lock manager implemented to control allocation of resources to processes, and to maintain a list of processes waiting for resources and a list of resources that are allocated to a process.

14. The computing system of claim 11 further including:

a hang manager configured to maintain a list of processes that are in a wait state; and

where the blocking avoidance logic is configured to access the hang manager to determine when a process is a waiting process.

15. The computing system of claim 11 where the blocking avoidance logic is further configured to:

identify the resource that the waiting process is waiting for by searching the shared global memory on the remote node;

based on the identity of the resource, determine the identity of the blocking process to which the resource is allocated.

16. The computing system of claim 15 where resource is identified by a resource identifier that was assigned to the resource.

17. The computing system of claim 11 where the blocking avoidance logic is configured to selectively cause the operating system to increase the priority by:

determining a node to which the blocking process is local; and

transmitting a message to the node including state information of the waiting process;

where the message is configured to cause the operating system to change the priority of the blocking process based on the state information of the waiting process and the execution priority of the blocking process, where the priority of the blocking process is increased when the blocking process is not executing and if the execution priority of the blocking process is lower than the execution priority of the waiting process.

18. The computing system of claim 11 where the blocking avoidance logic is configured to:

search process state information of the plurality of computing nodes to identify the waiting process and determine a resource identifier of the resource that the waiting process is waiting for; and

search resource state information of the plurality of computing nodes to identify the blocking process using the resource identifier.

19. The computing system of claim 11 where the global avoidance logic is configured to determine when the process is waiting by detecting a timeout by the process and search the process state information in the global shared memory for waiting processes that are blocked by another process.

20. The computing system of claim 11, wherein selectively causing an operating system local to the blocking process to increase the execution priority of the blocking process includes lending the execution priority of the waiting process to the blocking process and returning the lent execution priority of the waiting process when the blocking process releases the resource.

21. The computing system of claim 11, further comprising:

after the resource is released by the blocking process, passing the increased execution priority from the blocking process to a subsequent process that uses the resource, wherein the subsequent process is the waiting process.

22. A non-transitory computer-readable medium storing computer-executable instructions that when executed by a computer cause the computer to perform a method, the method comprising:

searching process state information of a distributed computing system having a plurality of computing nodes to identify a waiting process from a first computing node, where the waiting process is in a blocked state and waiting for a resource;

searching resource state information of the distributed computing system to identify a blocking process from a different computing node in the distributing computing system using an identifier of the resource, where the blocking process is holding the resource, where the process state information and the resource state information are stored in a global memory shared by the plurality of computing nodes, and where the global memory is on a remote node in the plurality of computing nodes and is accessed by the plurality of computing nodes;

generating a message that includes instructions and data with at least a priority of the waiting process, an identifier of the blocking process, and the identifier of the resource; and

transmitting the message to the second node to cause an execution priority of the blocking process to be increased if the execution priority is lower than the waiting process.

23. The non-transitory computer-readable medium of claim 22 further including:

determining the identifier associated with the resource; and

where identifying the blocking process is based on the resource identifier, where increasing an execution priority of the blocking process includes extending an execution time of the blocking process if the blocking process is currently executing and increasing an execution priority of the blocking process if the blocking process is not currently executing.

24. The non-transitory computer-readable medium of claim 22 where transmitting the message to the second node causes an operating system local to the second node to increase the execution priority of the blocking process if the execution priority is lower than the waiting process.

25. The non-transitory computer-readable medium of claim 22, wherein causing the execution priority of the blocking process to be increased includes lending the execution priority of the waiting process to the blocking process and returning the lent priority when the blocking process releases the resource.

26. The non-transitory computer-readable medium of claim 22, further comprising:

after the resource is released by the blocking process, passing the increased execution priority from the blocking process to a subsequent process that uses the resource.