From jayavant.patil82 at gmail.com Wed Jan 4 06:36:33 2012 From: jayavant.patil82 at gmail.com (Jayavant Patil) Date: Wed, 4 Jan 2012 19:06:33 +0530 Subject: [Mauiusers] USERCFG and GROUPCFG Throttling policies Message-ID: Hi, I am using TORQUE 3.0.0 and Maui 3.3. Has anybody ever used USERCFG, GROUPCFG and ACCOUNTCFG's throttling policies? When I was testing with these policy parameters like MAXNODE,MAXPROC etc., I found that they are not working. e.g. I specified the following in maui.cfg : 1. USERCFG[root] MAXNODE=1 GROUPCFG[root] MAXNODE=0 job submission from root user still runs 2. USERCFG[root] MAXNODE=0 GROUPCFG[root] MAXNODE=1 job submission from root user still runs 3. USERCFG[root] MAXNODE=0 GROUPCFG[root] MAXNODE=0 job submission from root user still runs 4. USERCFG[root] MAXNODE=0 job submission from root user still runs OR GROUPCFG[root] MAXNODE=0 job submission from root user still runs and the same result with MAXPROC parameter (but MAXJOB parameter working). My second question is what is the difference between USERCFG and ACCOUNTCFG and also between GROUPCFG and ACCOUNTCFG? Thanks in advance. -- Regards, Jayavant Ningoji Patil +91 9923536030. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20120104/86ea7c6d/attachment.html From jayavant.patil82 at gmail.com Mon Jan 9 03:27:55 2012 From: jayavant.patil82 at gmail.com (Jayavant Patil) Date: Mon, 9 Jan 2012 15:57:55 +0530 Subject: [Mauiusers] USERCFG and GROUPCFG Throttling policies Message-ID: Hi, I am using TORQUE 3.0.0 and Maui 3.3. Has anybody ever used USERCFG, GROUPCFG and ACCOUNTCFG's throttling policies? When I was testing with these policy parameters like MAXNODE,MAXPROC etc., I found that they are not working when I explicitly specify the nodes in the -l nodes= option. What could be the reasons? My second question is what is the difference between USERCFG and ACCOUNTCFG and also between GROUPCFG and ACCOUNTCFG? -- Thanks & Regards, Jayavant Ningoji Patil +91 9923536030. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20120109/a7c50b6b/attachment.html From jbllistes at gmail.com Tue Jan 10 13:23:07 2012 From: jbllistes at gmail.com (jbllistes) Date: Tue, 10 Jan 2012 21:23:07 +0100 Subject: [Mauiusers] Hybrid scenario - FairShare and FIFO Policy Message-ID: Hi, We would like to setup FairShare policy on our scheduler, but we will have an hybrid scenario mixing FairShare Policy for shared resources and FIFO policy for some queues or nodes. It will be something like the following diagram but with more departments: ** We would like to know how to do that. The key point is that the accumulated consumption of each group in their own machines (not shared resources, deptA and deptB in the diagram above) must not be taken into account in the shared FairSharing system (common resources). In other words, we want to separate our cluster into FairSharing resources and FIFO resources using the same MAUI server. On the other hand, we would like to know how to adjust the weight of the share with the relative performance of different kind of processors for the shared resources. We thought that SPEED option works for that but we had discovered that this is only used for allocation rules. We use Torque 2.5.8+Maui 3.3.1. Please, let us know if you need more information to help us. Thank you in advance. Best regards, Jordi & Carles -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20120110/23ef8edd/attachment.html From stefan.friedel at iwr.uni-heidelberg.de Thu Jan 26 00:46:29 2012 From: stefan.friedel at iwr.uni-heidelberg.de (Stefan Friedel) Date: Thu, 26 Jan 2012 08:46:29 +0100 Subject: [Mauiusers] procseconds utilized in maui stats wrong? Message-ID: <20120126074629.GA3107@woyzeck> Good morning, could somebody explain why the "ProcSeconds utilized" field in the maui stats files (e.g. /var/spool/maui/stats/Mon_Jun_23_2008) are obviously multiplied by the number of tasks? E.g.: ######### 5830 0 128 14400 Completed [helics2:1] 1214131474 1214228182 1214228182 1214238255 [NONE] [NONE] [NONE] >= 0M >= 0M [helics2] 1214131474 128 4 [NONE]:DEFAULT [RESTARTABLE] [NONE] [NONE] [NONE] 25 159218962.33 DEFAULT 1 0M 0M 2000M 0 2140000000 opt103:opt102:opt101:opt100:opt099:opt098:opt097:opt096:opt095:opt094:opt093:opt092:opt091:opt090:opt089:opt088:opt087:opt086:opt085:opt084:opt083:opt081:opt080:opt079:opt061:opt060:opt059:opt058:opt057:opt056:opt055:opt054 base [NONE] [NONE] [DEFAULT] [NONE] [NONE] ########## The job asked for 14400s=4h and 128 tasks (it runs on 32x4-core nodes=128 mpi processes), actually it runs for stop-start=1214238255s-1214228182s=10073s => the max. "ProcSeconds utilized" should be less or equal to 128*10073s=1289344s. But the value is 159218962.33s which is little bit less than 128^2*10073s...: 159218962.33s/128^2~=9800s. (I checked many jobs in many stat files from the last ~4 years: all of them had the same task^2 factor...) any hint/idea? Thanks and regards -- Stefan Friedel Zentrale Dienste - Interdisziplin?res Zentrum f?r Wissenschaftliches Rechnen der Universit?t Heidelberg - IWR - INF 368, 69120 Heidelberg stefan.friedel at iwr.uni-heidelberg.de ? +49 6221 54-8240 ? -5224 -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 198 bytes Desc: Digital signature Url : http://www.supercluster.org/pipermail/mauiusers/attachments/20120126/3d8d97a2/attachment-0001.bin