[jira] [Comment Edited] (YARN-2162) add ability in Fair Scheduler to optionally configure maxResources in terms of percentage
[ https://issues.apache.org/jira/browse/YARN-2162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16189056#comment-16189056 ] Yufei Gu edited comment on YARN-2162 at 10/2/17 11:35 PM: -- Run SLS with 400 nodes, 200 apps, and 2k containers. Both base line and patch YARN-2162 runs 20 times. Result is in uploaded file test-400nm-200app-2k_NODE_UPDATE.timecost.svg. Please ignore the "Resource Type 0", it is the yarn-2162 patch actually. No obvious performance regression. The patch version even performs a little better due to normal fluctuation. was (Author: yufeigu): Run SLS with 400 nodes, 200 apps, and 2k containers. Both base line and patch YARN-2162 runs 20 times. !test-400nm-200app-2k_NODE_UPDATE.timecost.svg|thumbnail!. Please ignore the "Resource Type 0", it is the yarn-2162 patch actually. Don't see obvious performance regression. The patch version even performs a little better due to normal fluctuation. > add ability in Fair Scheduler to optionally configure maxResources in terms > of percentage > - > > Key: YARN-2162 > URL: https://issues.apache.org/jira/browse/YARN-2162 > Project: Hadoop YARN > Issue Type: Improvement > Components: fairscheduler, scheduler >Reporter: Ashwin Shankar >Assignee: Yufei Gu > Labels: scheduler > Attachments: test-400nm-200app-2k_NODE_UPDATE.timecost.svg, > YARN-2162.001.patch, YARN-2162.002.patch, YARN-2162.003.patch, > YARN-2162.004.patch, YARN-2162.005.patch, YARN-2162.006.patch, > YARN-2162.007.patch > > > minResources and maxResources in fair scheduler configs are expressed in > terms of absolute numbers X mb, Y vcores. > As a result, when we expand or shrink our hadoop cluster, we need to > recalculate and change minResources/maxResources accordingly, which is pretty > inconvenient. > We can circumvent this problem if we can optionally configure these > properties in terms of percentage of cluster capacity. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - 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-2162) add ability in Fair Scheduler to optionally configure maxResources in terms of percentage
[ https://issues.apache.org/jira/browse/YARN-2162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16189056#comment-16189056 ] Yufei Gu edited comment on YARN-2162 at 10/2/17 11:34 PM: -- Run SLS with 400 nodes, 200 apps, and 2k containers. Both base line and patch YARN-2162 runs 20 times. !test-400nm-200app-2k_NODE_UPDATE.timecost.svg|thumbnail!. Please ignore the "Resource Type 0", it is the yarn-2162 patch actually. Don't see obvious performance regression. The patch version even performs a little better due to normal fluctuation. was (Author: yufeigu): Run SLS with 400 nodes, 200 apps, and 2k containers. Both base line and patch YARN-2162 runs 20 times. !attachment-name.jpg|thumbnail!. Please ignore the "Resource Type 0", it is the yarn-2162 patch actually. Don't see obvious performance regression. The patch version even performs a little better due to normal fluctuation. > add ability in Fair Scheduler to optionally configure maxResources in terms > of percentage > - > > Key: YARN-2162 > URL: https://issues.apache.org/jira/browse/YARN-2162 > Project: Hadoop YARN > Issue Type: Improvement > Components: fairscheduler, scheduler >Reporter: Ashwin Shankar >Assignee: Yufei Gu > Labels: scheduler > Attachments: test-400nm-200app-2k_NODE_UPDATE.timecost.svg, > YARN-2162.001.patch, YARN-2162.002.patch, YARN-2162.003.patch, > YARN-2162.004.patch, YARN-2162.005.patch, YARN-2162.006.patch, > YARN-2162.007.patch > > > minResources and maxResources in fair scheduler configs are expressed in > terms of absolute numbers X mb, Y vcores. > As a result, when we expand or shrink our hadoop cluster, we need to > recalculate and change minResources/maxResources accordingly, which is pretty > inconvenient. > We can circumvent this problem if we can optionally configure these > properties in terms of percentage of cluster capacity. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - 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-2162) add ability in Fair Scheduler to optionally configure maxResources in terms of percentage
[ https://issues.apache.org/jira/browse/YARN-2162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16178690#comment-16178690 ] Yufei Gu edited comment on YARN-2162 at 9/25/17 7:59 AM: - Thanks for the review, [~templedf]. Uploaded patch v6 for your comments. - 1. Add the check in class {{AllocationFileLoaderService}} . - 2. Deleted - 3. Done except for policy. - 4. FairScheduler#getResourceCalculator() returns null because FairScheduler is mocked. - 5. What I am thinking is using 'cpu' is better than using 'vcores'. We could consider 'vcores' as the unit of cpu just as 'mb' is the unit of memory. Besides, it is a little confusing between 2 vcores and 50% vcores. Is 50% vcores equals to 0.5 vcores? - 6. Cannot drop last white space since we want to support case like this "50% memory"; it may be more readable if we have that parentheses; make it a '*' since we wanna support case like "50. % memory". - 7. Add the negative tests. - 8. Add tests for class ConfigurableResource - 9. Let's try some perf tests on SLS later. was (Author: yufeigu): Thanks for the review, [~templedf]. - 1. Add the check in class {{AllocationFileLoaderService}} . - 2. Deleted - 3. Done except for policy. - 4. FairScheduler#getResourceCalculator() returns null because FairScheduler is mocked. - 5. What I am thinking is using 'cpu' is better than using 'vcores'. We could consider 'vcores' as the unit of cpu just as 'mb' is the unit of memory. Besides, it is a little confusing between 2 vcores and 50% vcores. Is 50% vcores equals to 0.5 vcores? - 6. Cannot drop last white space since we want to support case like this "50% memory"; it may be more readable if we have that parentheses; make it a '*' since we wanna support case like "50. % memory". - 7. Add the negative tests. - 8. Add tests for class ConfigurableResource - 9. Let's try some perf tests on SLS later. > add ability in Fair Scheduler to optionally configure maxResources in terms > of percentage > - > > Key: YARN-2162 > URL: https://issues.apache.org/jira/browse/YARN-2162 > Project: Hadoop YARN > Issue Type: Improvement > Components: fairscheduler, scheduler >Reporter: Ashwin Shankar >Assignee: Yufei Gu > Labels: scheduler > Attachments: YARN-2162.001.patch, YARN-2162.002.patch, > YARN-2162.003.patch, YARN-2162.004.patch, YARN-2162.005.patch, > YARN-2162.006.patch > > > minResources and maxResources in fair scheduler configs are expressed in > terms of absolute numbers X mb, Y vcores. > As a result, when we expand or shrink our hadoop cluster, we need to > recalculate and change minResources/maxResources accordingly, which is pretty > inconvenient. > We can circumvent this problem if we can optionally configure these > properties in terms of percentage of cluster capacity. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org