[jira] [Comment Edited] (YARN-6428) Queue AM limit is not honored in CS always
[ https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16036779#comment-16036779 ] Bibin A Chundatt edited comment on YARN-6428 at 6/5/17 10:28 AM: - During offline discussion with [~naganarasimha...@apache.org] for {code} @Override public Resource multiplyAndNormalizeUp(Resource r, double by, Resource stepFactor) { return Resources.createResource( roundUp( (int)Math.ceil(r.getMemorySize() * by), stepFactor.getMemorySize()), roundUp( (int)Math.ceil(r.getVirtualCores() * by), stepFactor.getVirtualCores()) ); } {code} In the above base code instead off {{(int)Math.ceil(r.getMemorySize() * by)}} it should be {{(long)Math.ceil(r.getMemorySize() * by)}} . For *DefaultResourceCalculator* {{long}} is used but *DominantResourceCalculator* {{int}} is used.Seems to be a miss during type datatype change was (Author: bibinchundatt): Adding point we had in offline discussion with [~naganarasimha...@apache.org] {code} @Override public Resource multiplyAndNormalizeUp(Resource r, double by, Resource stepFactor) { return Resources.createResource( roundUp( (int)Math.ceil(r.getMemorySize() * by), stepFactor.getMemorySize()), roundUp( (int)Math.ceil(r.getVirtualCores() * by), stepFactor.getVirtualCores()) ); } {code} In the above base code instead off {{(int)Math.ceil(r.getMemorySize() * by)}} it should be {{(long)Math.ceil(r.getMemorySize() * by)}} . For DefaultResource Calculator {{long}} is used but *DefaultResourceCalculator* {{int}} is used. > Queue AM limit is not honored in CS always > --- > > Key: YARN-6428 > URL: https://issues.apache.org/jira/browse/YARN-6428 > Project: Hadoop YARN > Issue Type: Bug >Reporter: Bibin A Chundatt >Assignee: Bibin A Chundatt > Attachments: YARN-6428.0001.patch > > > Steps to reproduce > > Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each. > Configure 100% to default queue as capacity and max am limit as 10 % > Minimum scheduler memory and vcore as 512,1 > *Expected* > AM limit 4096 and 4 vores > *Actual* > AM limit 4096+512 and 4+1 vcore -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Comment Edited] (YARN-6428) Queue AM limit is not honored in CS always
[ https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15978691#comment-15978691 ] Bibin A Chundatt edited comment on YARN-6428 at 4/25/17 4:43 PM: - [~sunilg] / [~leftnoteasy] Could you please share your thoughts.? was (Author: bibinchundatt): [~sunilg] / [~leftnoteasy] Any comments? > Queue AM limit is not honored in CS always > --- > > Key: YARN-6428 > URL: https://issues.apache.org/jira/browse/YARN-6428 > Project: Hadoop YARN > Issue Type: Bug >Reporter: Bibin A Chundatt >Assignee: Bibin A Chundatt > Attachments: YARN-6428.0001.patch > > > Steps to reproduce > > Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each. > Configure 100% to default queue as capacity and max am limit as 10 % > Minimum scheduler memory and vcore as 512,1 > *Expected* > AM limit 4096 and 4 vores > *Actual* > AM limit 4096+512 and 4+1 vcore -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Comment Edited] (YARN-6428) Queue AM limit is not honored in CS always
[ https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15953940#comment-15953940 ] Bibin A Chundatt edited comment on YARN-6428 at 4/3/17 6:18 PM: {{DefaultResourceCalculator}} any idea why {{0.5}} is added don't find the same in {{DominantResourceCalculator}} {code} @Override public Resource multiplyAndNormalizeUp(Resource r, double by, Resource stepFactor) { return Resources.createResource( roundUp((long) (r.getMemorySize() * by + 0.5), stepFactor.getMemorySize())); } {code} We should use {{Math.ceil}} here too. was (Author: bibinchundatt): {{DefaultResourceCalculator}} any idea why {{0.5}} is added don't find the same in {{DominantResourceCalculator}} {code} @Override public Resource multiplyAndNormalizeUp(Resource r, double by, Resource stepFactor) { return Resources.createResource( roundUp((long) (r.getMemorySize() * by + 0.5), stepFactor.getMemorySize())); } {code} > Queue AM limit is not honored in CS always > --- > > Key: YARN-6428 > URL: https://issues.apache.org/jira/browse/YARN-6428 > Project: Hadoop YARN > Issue Type: Bug >Reporter: Bibin A Chundatt >Assignee: Bibin A Chundatt > Attachments: YARN-6428.0001.patch > > > Steps to reproduce > > Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each. > Configure 100% to default queue as capacity and max am limit as 10 % > Minimum scheduler memory and vcore as 512,1 > *Expected* > AM limit 4096 and 4 vores > *Actual* > AM limit 4096+512 and 4+1 vcore -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Comment Edited] (YARN-6428) Queue AM limit is not honored in CS always
[ https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15953068#comment-15953068 ] Bibin A Chundatt edited comment on YARN-6428 at 4/3/17 12:44 PM: - {{ResourceCalculator#multiplyAndNormalizeUp}} using double to multiply and ceil for am limit factor. {code} roundUp( (int)Math.ceil(r.getMemorySize() * by) {code} {{0.1}} will be {{0.1012312444}} so multiple and ceil and round by will give give 4097 instead of 4096. was (Author: bibinchundatt): {{ResourceCalculator#multiplyAndNormalizeUp}} using double to multiply and ceil for am limit factor. {code} roundUp( (int)Math.ceil(r.getMemorySize() * by) {code} will give 4097 instead of 4096 . > Queue AM limit is not honored in CS always > --- > > Key: YARN-6428 > URL: https://issues.apache.org/jira/browse/YARN-6428 > Project: Hadoop YARN > Issue Type: Bug >Reporter: Bibin A Chundatt >Assignee: Bibin A Chundatt > Attachments: YARN-6428.0001.patch > > > Steps to reproduce > > Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each. > Configure 100% to default queue as capacity and max am limit as 10 % > Minimum scheduler memory and vcore as 512,1 > *Expected* > AM limit 4096 and 4 vores > *Actual* > AM limit 4096+512 and 4+1 vcore -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Comment Edited] (YARN-6428) Queue AM limit is not honored in CS always
[ https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15953405#comment-15953405 ] Bibin A Chundatt edited comment on YARN-6428 at 4/3/17 12:43 PM: - [~varun_saxena] Thank you for looking into the issue. The second approach will be a bigger change than current. for example: {{QueueCapacities}} currently uses float . In current implementation we have option to used both interface. was (Author: bibinchundatt): [~varun_saxena] Thank you for looking into the issue. The second approach will be a bigger change that current. for example: {{QueueCapacities}} currently uses float . In current implementation we have option to used both interface. > Queue AM limit is not honored in CS always > --- > > Key: YARN-6428 > URL: https://issues.apache.org/jira/browse/YARN-6428 > Project: Hadoop YARN > Issue Type: Bug >Reporter: Bibin A Chundatt >Assignee: Bibin A Chundatt > Attachments: YARN-6428.0001.patch > > > Steps to reproduce > > Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each. > Configure 100% to default queue as capacity and max am limit as 10 % > Minimum scheduler memory and vcore as 512,1 > *Expected* > AM limit 4096 and 4 vores > *Actual* > AM limit 4096+512 and 4+1 vcore -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org