[jira] [Commented] (SCB-353) Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义
[ https://issues.apache.org/jira/browse/SCB-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16478494#comment-16478494 ] wujimin commented on SCB-353: - nobody can get p90/p99 except we record times of each invocation i'm not sure what the last decision. > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义 > --- > > Key: SCB-353 > URL: https://issues.apache.org/jira/browse/SCB-353 > Project: Apache ServiceComb > Issue Type: Task > Components: Java-Chassis >Affects Versions: java-chassis-1.0.0-m2 >Reporter: renyingxin >Assignee: wujimin >Priority: Major > Fix For: java-chassis-1.0.0-m2 > > > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义,也就是类似[0,200),[200,400)这种时延分布的统计,平均时延可能会掩盖系统的部分问题,比如部分高延时,但是被正常平均值掩盖了 -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (SCB-353) Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义
[ https://issues.apache.org/jira/browse/SCB-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16478473#comment-16478473 ] Willem Jiang commented on SCB-353: -- [~wujimin] If we don't provide this feature, we can add some comments and close the JIRA for it. BTW, it looks like user can get the p90/p99 by using monitor backend. > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义 > --- > > Key: SCB-353 > URL: https://issues.apache.org/jira/browse/SCB-353 > Project: Apache ServiceComb > Issue Type: Task > Components: Java-Chassis >Affects Versions: java-chassis-1.0.0-m2 >Reporter: renyingxin >Assignee: wujimin >Priority: Major > Fix For: java-chassis-1.0.0-m2 > > > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义,也就是类似[0,200),[200,400)这种时延分布的统计,平均时延可能会掩盖系统的部分问题,比如部分高延时,但是被正常平均值掩盖了 -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (SCB-353) Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义
[ https://issues.apache.org/jira/browse/SCB-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16478367#comment-16478367 ] wujimin commented on SCB-353: - no, infact what customer want is not buckets, but p90/p99 and so on but implement buckets almost will not affect performance and implement pxx need to save times of each invocation, and sort them, will make performance bad. > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义 > --- > > Key: SCB-353 > URL: https://issues.apache.org/jira/browse/SCB-353 > Project: Apache ServiceComb > Issue Type: Task > Components: Java-Chassis >Affects Versions: java-chassis-1.0.0-m2 >Reporter: renyingxin >Assignee: wujimin >Priority: Major > Fix For: java-chassis-1.0.0-m2 > > > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义,也就是类似[0,200),[200,400)这种时延分布的统计,平均时延可能会掩盖系统的部分问题,比如部分高延时,但是被正常平均值掩盖了 -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (SCB-353) Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义
[ https://issues.apache.org/jira/browse/SCB-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16478309#comment-16478309 ] Willem Jiang commented on SCB-353: -- [~wujimin] Any update for this issue? Did we address this issue by refactoring the metrics? > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义 > --- > > Key: SCB-353 > URL: https://issues.apache.org/jira/browse/SCB-353 > Project: Apache ServiceComb > Issue Type: Task > Components: Java-Chassis >Affects Versions: java-chassis-1.0.0-m2 >Reporter: renyingxin >Assignee: wujimin >Priority: Major > Fix For: java-chassis-1.0.0-m2 > > > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义,也就是类似[0,200),[200,400)这种时延分布的统计,平均时延可能会掩盖系统的部分问题,比如部分高延时,但是被正常平均值掩盖了 -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (SCB-353) Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义
[ https://issues.apache.org/jira/browse/SCB-353?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16379941#comment-16379941 ] yangyongzheng commented on SCB-353: --- We will add a new tag for latency named **bucket** If user set **servicecomb.metrics.invocation.latency.addition = bucket**, monitor registry will auto register a set of bucket latency counter like: **servicecomb.metrics.invocation.latency.buckets= 50,100,500** **servicecomb.metrics.invocation.latency.unit= MILLISECONDS** servicecomb.invocation(operation=calculator.calculatorRestEndpoint.calculate,role=producer,stage=total,statistic=latency,status=200,unit=MILLISECONDS) servicecomb.invocation(operation=calculator.calculatorRestEndpoint.calculate,role=producer,bucket=050ms,stage=total,statistic=count,status=200,unit=MILLISECONDS) servicecomb.invocation(operation=calculator.calculatorRestEndpoint.calculate,role=producer,bucket=100ms,stage=total,statistic=count,status=200,unit=MILLISECONDS) servicecomb.invocation(operation=calculator.calculatorRestEndpoint.calculate,role=producer,bucket=500ms,stage=total,statistic=count,status=200,unit=MILLISECONDS) servicecomb.invocation(operation=calculator.calculatorRestEndpoint.calculate,role=producer,bucket=overflow,stage=total,statistic=count,status=200,unit=MILLISECONDS) All stage _queue,execution,total_ and all operation will be affected > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义 > --- > > Key: SCB-353 > URL: https://issues.apache.org/jira/browse/SCB-353 > Project: Apache ServiceComb > Issue Type: Task > Components: Java-Chassis >Affects Versions: java-chassis-1.0.0-m2 >Reporter: renyingxin >Assignee: yangyongzheng >Priority: Major > Fix For: java-chassis-1.0.0-m2 > > > Servicecomb当前性能统计不支持时延分布的统计,这个对于业务有重要意义,也就是类似[0,200),[200,400)这种时延分布的统计,平均时延可能会掩盖系统的部分问题,比如部分高延时,但是被正常平均值掩盖了 -- This message was sent by Atlassian JIRA (v7.6.3#76005)