Ticket 13276

Summary: PMI plugin modification
Product: Slurm Reporter: daiyq <daiyq>
Component: PMIxAssignee: Tim Wickberg <tim>
Status: OPEN --- QA Contact:
Severity: C - Contributions    
Priority: ---    
Version: 20.11.7   
Hardware: Linux   
OS: Linux   
Site: -Other- Alineos Sites: ---
Atos/Eviden Sites: --- Confidential Site: ---
Coreweave sites: --- Cray Sites: ---
DS9 clusters: --- HPCnow Sites: ---
HPE Sites: --- IBM Sites: ---
NOAA SIte: --- NoveTech Sites: ---
Nvidia HWinf-CS Sites: --- OCF Sites: ---
Recursion Pharma Sites: --- SFW Sites: ---
SNIC sites: --- Tzag Elita Sites: ---
Linux Distro: --- Machine Name:
CLE Version: Version Fixed:
Target Release: --- DevPrio: ---
Emory-Cloud Sites: ---
Attachments: patch

Description daiyq@nudt.edu.cn 2022-01-27 07:14:35 MST 
Created attachment 23170 [details]
patch

Some modifications regarding PMI2 and PMIx plugins.These changes, mainly done by Hongjia Cao (the original author of the PMI2 plug-in) and Yiqin Dai, reduce the PMI2 plug-in failures when running large scale jobs and reduce the time overhead steps in the PMIx plug-in. One of the changes to the PMI2 plug-in is proposed to be added to the official release.


This is a problem found by the National University of Defense Technology when applying slurm to large scale supercomputers (over 10^6 cores), and we classify the problem into five categories, of which the above problem is one.We provide a simpler patch and look forward to communicating with and contributing code to the Slurm community.  The authors of this patch are Hongjia Cao and Yiqin Dai at daiyq@nudt.edu.cn.