OpenPMIx
1. Quick start
2. Getting help
3. Release notes
4. Exceptions to the PMIx Standard
5. Building and installing PMIx
6. How Things Work
6.1. Session Directories
6.2. Client Init/Finalize Lifecycle and State Management
6.3. Distance Computation
6.4. Scheduler Integration
6.5. Process Sets and Groups
6.5.1. Overview
6.5.2. Process Sets
6.5.3. Group Construction, Destruction, and Fault Tolerance
6.5.4. Group Operations: Theory and Implementation
6.6. Server-Side Collective Tracking
6.7. Modex: Exchanging Process Data
6.8. The Shared-Memory Datastore:
gds/shmem3
6.9. Allocation Inheritance
6.10. Output Forwarding for Spawned Jobs
6.11. Resolving Peers and Nodes
6.12. How to Add a PMIx Datatype
6.13. PMIx_Log Support
6.14. Regular Expression (Node/Proc Map) Support
6.15. Transport Layer (Client/Server/Tool Connections)
6.16. Process & Resource Statistics (Monitoring)
6.17. Non-Blocking Operations in the Python Bindings
7. Release Notes
8. History
9. Version Numbers and Binary Compatibility
10. The Modular Component Architecture (MCA)
11. Building PMIx applications
12. Developer’s guide
13. Contributing to OpenPMIx
14. Known gaps and deferred work
15. Deep review notes
16. OpenPMIx Community Code of Conduct
17. License
18. OpenPMIx Security Policy
19. News
20. OpenPMIx manual pages
OpenPMIx
6.
How Things Work
6.5.
Process Sets and Groups
View page source
6.5.
Process Sets and Groups
This section describes PMIx support for process sets and groups.
6.5.1. Overview
6.5.1.1. Process Sets vs. Groups
6.5.1.2. The Group Lifecycle
6.5.1.3. Fault Tolerance
6.5.2. Process Sets
6.5.2.1. Process Set Attributes
6.5.3. Group Construction, Destruction, and Fault Tolerance
6.5.3.1. Collective Method
6.5.3.2. Bootstrap Method
6.5.3.3. Invite Method
6.5.3.4. Group Destruction
6.5.3.5. Leaving a Group
6.5.3.6. Fault Tolerance
6.5.4. Group Operations: Theory and Implementation
6.5.4.1. Two realizations of “group construction”
6.5.4.2. Client side
6.5.4.3. Server side
6.5.4.4. Fault tolerance: identity-based accounting
6.5.4.5. The host callback boundary
6.5.4.6. Event and attribute reference
6.5.4.7. Testing