Method and apparatus for handling failures of resource managers in a clustered environment

• US 6,671,704 B1
• Filed: 03/11/1999
• Issued: 12/30/2003
• Est. Priority Date: 03/11/1999
• Status: Expired due to Fees

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Lanka, Sitaram V., de Roo, John S., Sripada, Ramprasad K. L., Klein, Johannes, Gondi, Albert C.
• Primary Examiner(s): Le, Uyen
• Assistant Examiner(s): CHEN, TE Y
• Application Number:
  US09/267,032
• Time in Patent Office: 1,755 Days
• Field of Search: 707/1-10, 707/200, 707/511, 707/201-204, 709/101, 709/248, 709/227, 709/201, 709/237, 714/20, 714/16, 714/2, 714/10-13, 714/4
• US Class Current: 707/648
• CPC Class Codes: G06F 11/006 Identification G06F11/2289 takes precedence
• CPC Class Codes: G06F 11/2028 eliminating a faulty processor or activating a spare
• CPC Class Codes: G06F 11/2035 without idle spare hardware
• CPC Class Codes: G06F 11/2046 where the redundant components share persistent storage G06F11/2043 takes precedence
• CPC Class Codes: Y10S 707/99953 Recoverability
• CPC Class Codes: Y10S 707/99955 Archiving or backup

Abstract

A transaction processing system manages an information collection by employing a two-phase transaction commit protocol. Transaction participants are provided a unique identifier for use in communicating with the transaction management facility in a distributed environment. If the participant is lost, a backup participant is created, and provided with the same identifier as the failed participant to carry on with the transaction in place of the failed participant.

Claims

1. A method of providing fault tolerant operation of a transaction processing system, the method comprising the steps of:

sending a prepare signal to a resource manager process to begin the process of committing a transaction, the resource manager process being a participant in the transaction and identified by a globally unique identifier;

receiving a ready signal from the resource manager in response to the prepare signal, the ready signal being accompanied by state information regarding the state of the resource manager process; and

subsequent to the receipt of the ready signal,

storing the received state information;

receiving notice of a loss of the resource manager process;

broadcasting a commit signal in response to the ready signal;

creating a backup resource manager process for taking over the function and operation of the lost resource manager process, the lost resource manager process being unable to receive or respond to the commit signal;

assigning to the backup resource manager process the globally unique identifier of the lost resource manager process; and

causing the backup resource manager process to complete the process of committing the transaction by (i) sending the stored state information to the backup resource manager process, in response to a query by the backup resource manager process using the assigned identifier, the backup resource manager process determining, from the stored state information, that the lost resource manager was a participant in the transaction, and (ii) sending information regarding the state of the transaction to the backup resource manager process, in response to a query by the backup resource manager process, the backup resource manager process taking steps to commit the transaction based on the transaction state received from the query.

2. A method as recited in claim 1,

wherein the transaction processing system includes at least two central processing units (CPUs);

wherein the resource manager includes a local resource manager component operating in each CPU; and

wherein one of the local resource managers is a beginner resource manager component.

3. A method as recited in claim 2, further comprising the step of, prior to sending the prepare signal, creating the resource manager process by

providing the resource manager with various interfaces for communicating with the local resource manager component,

assigning the identifier to the resource manager in order to identify the resource manager to the local resource manager component, and

responding to a query from the resource manager with information indicating any transactions the resource manager should be aware of.

4. A method as recited in claim 2, further comprising the step of, prior to sending a prepare signal, registering by the local resource manager the resource manager as a participant in the transaction.

5. A method as recited in claim 2,

wherein each CPU has resident thereon a local distributed transaction coordinator gateway;

and

wherein the resource manager communicates with the local resource manager component via the local distributed transaction coordinator gateway.

6. A method as recited in claim 2, wherein the step of storing the state information includes storing the state information in an audit log by the local resource manager component.

7. A method as recited in claim 2, wherein the step of creating the backup resource manager includes providing the created resource manager with various interfaces for communicating with the local resource manager component.

8. A method as recited in claim 2,

wherein each CPU has a cluster service program resident thereon;

and

wherein the step of creating a backup resource manager is performed by one of the cluster service programs.

9. A fault tolerant transaction processing system comprising:

a communication fabric for providing interprocessor and input/output communication;

a plurality of peripheral devices connected to the communication fabric, for storing a database thereon;

a plurality of central processor units (CPUs), connected to the communication fabric to communicate with the plurality of peripheral devices,

at least one of the CPUs having a resource manager process having a globally unique identifier, and configured to participate in a transaction, modify the database, and report information regarding its state in a ready signal, and

each CPU having a cluster service component configured to detect the loss of a process, including a resource manager process, create a backup resource manager process for taking over the function and operation of the lost resource manager process, including the operation of committing a transaction, and assign to a backup resource manager process a globally unique identifier of a lost resource manager process,

the plurality of CPUs hosting a transaction manager facility (TMF) that includes a TMF component allocated to each CPU and configured to

send a prepare signal to any participating resource manager to begin the process of committing the transaction,

receive a ready signal from each resource manager participating in the transaction, the ready signal being accompanied by state information regarding the state of the resource manager process, and store the received state information,

receive notice of a loss of a resource manager process,

broadcast a commit to all of the CPUs having participating resource managers,

send the stored state information to backup resource manager process, in response to a query by the backup resource manager process using an assigned identifier, the backup resource manager process determining, from the stored state information, that the lost resource manager was a participant in the transaction, and

send information regarding the state of the transaction to a backup resource manager process, in response to a query by the backup resource manager process, the backup resource manager process taking steps to commit the transaction based on the transaction state received from the query.

10. A fault tolerant transaction processing system as recited in claim 9, further comprising a transaction manager process resident on one of the CPUs and in communication with the TMF.

11. A fault tolerant transaction processing system as recited in claim 9, further comprising a gateway component running on at least one of the CPUs, the gateway component for facilitating communications between the CPUs over the communications fabric.

12. A fault tolerant transaction processing system comprising:

communications means for providing interprocessor and input/output communication;

a plurality of peripheral devices connected to the communication fabric, for storing a database thereon;

a plurality of processing means, each connected to the communication fabric to communicate with the plurality of peripheral devices, the plurality of processing means including

at least one resource manager means, having a globally unique identifier, for participating in a transaction, modifying the database and reporting information regarding its state in a ready signal,

at least one gateway means for facilitating communications between the plurality of processing means over the communications means,

a cluster service means for detecting the loss of resource manager means, creating backup resource manager means for taking over the function and operation of the lost resource manager means, including the operation of committing a transaction, and assigning to backup resource manager means the globally unique identifier of a lost resource manager means, and

a transaction facility means for sending a prepare signal to any participating resource manager means to begin the process of committing the transaction, receiving a ready signal from each resource manager participating in the transaction, the ready signal being accompanied by state information regarding the state of the resource manager means, and storing the received state information, receiving notice of a loss of a resource manager means, broadcasting a commit to all processing means having participating resource manager means, sending the stored state information to a backup resource manager means, in response to a query by a backup resource manager means using the assigned globally unique identifier, the backup resource manager means determining, from the stored state information, that the lost resource manager means was a participant in the transaction, and sending information regarding the state of the transaction to a backup resource manager means, in response to a query by the backup resource manager process, the backup resource manager means taking steps to commit the transaction based on the transaction state received from the query.