[jira] [Created] (YARN-9509) Capped cpu usage with cgroup strict-resource-usage based on a mulitplier
Nicolas Fraison created YARN-9509: - Summary: Capped cpu usage with cgroup strict-resource-usage based on a mulitplier Key: YARN-9509 URL: https://issues.apache.org/jira/browse/YARN-9509 Project: Hadoop YARN Issue Type: New Feature Components: nodemanager Reporter: Nicolas Fraison Add a multiplier configuration on strict resource usage to authorize container to use spare cpu up to a limit. Currently with strict resource usage you can't get more than what you request which is sometime not good for jobs that doesn't have a constant usage of cpu (for ex. spark jobs with multiple stages). But without strict resource usage we have seen some bad behaviour from our users that don't tune at all their needs and it leads to some containers requesting 2 vcore but constantly using 20. The idea here is to still authorize containers to get more cpu than what they request if some are free but also to avoid too big differencies so SLA on jobs is not breached if the cluster is full (at least increase of runtime is contain) -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Commented] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
[ https://issues.apache.org/jira/browse/YARN-7793?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16365198#comment-16365198 ] Nicolas Fraison commented on YARN-7793: --- [~vinodkv] and [~rajesh.balamohan] could you please review. Thks > Lots of parsing error reported on nodemanager after activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > - > > Key: YARN-7793 > URL: https://issues.apache.org/jira/browse/YARN-7793 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7793.001.patch, YARN-7793.002.patch, YARN-7793.patch > > > After activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > nodemanagers are reporting lots of parsing error like this one > {code:java} > Error in parsing : INVALID : valuerd wr mr mw me ac{code} > Looking at smaps file I can see that there are in fact one line with no > Number: > {code:java} > VmFlags: rd ex mr mw me de{code} > This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Updated] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
[ https://issues.apache.org/jira/browse/YARN-7793?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Fraison updated YARN-7793: -- Attachment: YARN-7793.002.patch > Lots of parsing error reported on nodemanager after activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > - > > Key: YARN-7793 > URL: https://issues.apache.org/jira/browse/YARN-7793 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7793.001.patch, YARN-7793.002.patch, YARN-7793.patch > > > After activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > nodemanagers are reporting lots of parsing error like this one > {code:java} > Error in parsing : INVALID : valuerd wr mr mw me ac{code} > Looking at smaps file I can see that there are in fact one line with no > Number: > {code:java} > VmFlags: rd ex mr mw me de{code} > This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Updated] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
[ https://issues.apache.org/jira/browse/YARN-7793?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Fraison updated YARN-7793: -- Attachment: YARN-7793.001.patch > Lots of parsing error reported on nodemanager after activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > - > > Key: YARN-7793 > URL: https://issues.apache.org/jira/browse/YARN-7793 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7793.001.patch, YARN-7793.patch > > > After activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > nodemanagers are reporting lots of parsing error like this one > {code:java} > Error in parsing : INVALID : valuerd wr mr mw me ac{code} > Looking at smaps file I can see that there are in fact one line with no > Number: > {code:java} > VmFlags: rd ex mr mw me de{code} > This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Updated] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
[ https://issues.apache.org/jira/browse/YARN-7793?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Fraison updated YARN-7793: -- Attachment: YARN-7793.patch > Lots of parsing error reported on nodemanager after activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > - > > Key: YARN-7793 > URL: https://issues.apache.org/jira/browse/YARN-7793 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7793.patch > > > After activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > nodemanagers are reporting lots of parsing error like this one > {code:java} > Error in parsing : INVALID : valuerd wr mr mw me ac{code} > Looking at smaps file I can see that there are in fact one line with no > Number: > {code:java} > VmFlags: rd ex mr mw me de{code} > This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Updated] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
[ https://issues.apache.org/jira/browse/YARN-7793?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Fraison updated YARN-7793: -- Description: After activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled nodemanagers are reporting lots of parsing error like this one {code:java} Error in parsing : INVALID : valuerd wr mr mw me ac{code} Looking at smaps file I can see that there are in fact one line with no Number: {code:java} VmFlags: rd ex mr mw me de{code} This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree was: After activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled nodemanagers are reporting lots of parsing error like this one `Error in parsing : INVALID : valuerd wr mr mw me ac` Looking at smaps file I can see that there are in fact one line with no Number: `VmFlags: rd ex mr mw me de` which is retreived by the MEM_INFO_PATTERN > Lots of parsing error reported on nodemanager after activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > - > > Key: YARN-7793 > URL: https://issues.apache.org/jira/browse/YARN-7793 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > > After activating > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled > nodemanagers are reporting lots of parsing error like this one > {code:java} > Error in parsing : INVALID : valuerd wr mr mw me ac{code} > Looking at smaps file I can see that there are in fact one line with no > Number: > {code:java} > VmFlags: rd ex mr mw me de{code} > This line is retreived by the MEM_INFO_PATTERN in ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Created] (YARN-7793) Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled
Nicolas Fraison created YARN-7793: - Summary: Lots of parsing error reported on nodemanager after activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled Key: YARN-7793 URL: https://issues.apache.org/jira/browse/YARN-7793 Project: Hadoop YARN Issue Type: Bug Components: nodemanager Environment: cloudera release cdh-5.11.0 openjdk version "1.8.0_91" linux centos7 servers Reporter: Nicolas Fraison After activating yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled nodemanagers are reporting lots of parsing error like this one `Error in parsing : INVALID : valuerd wr mr mw me ac` Looking at smaps file I can see that there are in fact one line with no Number: `VmFlags: rd ex mr mw me de` which is retreived by the MEM_INFO_PATTERN -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Updated] (YARN-7772) yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not taken if process-tree class is not set
[ https://issues.apache.org/jira/browse/YARN-7772?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Fraison updated YARN-7772: -- Environment: cloudera release cdh-5.11.0 openjdk version "1.8.0_91" linux centos7 servers was: cloudera release cdh-5.11.0 openjdk version "1.8.0_91" linux centos6 servers > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not > taken if process-tree class is not set > > > Key: YARN-7772 > URL: https://issues.apache.org/jira/browse/YARN-7772 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos7 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7772.patch > > > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not > taken in account by ProcfsBasedProcessTree if > yarn.nodemanager.container-monitor.process-tree.class is not set in the > configuration. > The parameter is well taken in account if we set > yarn.nodemanager.container-monitor.process-tree.class to > org.apache.hadoop.yarn.util.ProcfsBasedProcessTree. > This is due to the getResourceCalculatorProcessTree which fallback to os > specific class if no ResourceCalculatorProcessTree class is provided and in > the fallback we never call setConf for ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Commented] (YARN-7772) yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not taken if process-tree class is not set
[ https://issues.apache.org/jira/browse/YARN-7772?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16335522#comment-16335522 ] Nicolas Fraison commented on YARN-7772: --- [~vinodkv] and [~rajesh.balamohan] could you please review. Thks > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not > taken if process-tree class is not set > > > Key: YARN-7772 > URL: https://issues.apache.org/jira/browse/YARN-7772 > Project: Hadoop YARN > Issue Type: Bug > Components: nodemanager > Environment: cloudera release cdh-5.11.0 > openjdk version "1.8.0_91" > linux centos6 servers >Reporter: Nicolas Fraison >Priority: Minor > Attachments: YARN-7772.patch > > > yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled is not > taken in account by ProcfsBasedProcessTree if > yarn.nodemanager.container-monitor.process-tree.class is not set in the > configuration. > The parameter is well taken in account if we set > yarn.nodemanager.container-monitor.process-tree.class to > org.apache.hadoop.yarn.util.ProcfsBasedProcessTree. > This is due to the getResourceCalculatorProcessTree which fallback to os > specific class if no ResourceCalculatorProcessTree class is provided and in > the fallback we never call setConf for ProcfsBasedProcessTree -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org