[ https://issues.apache.org/jira/browse/HDFS-13674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16533248#comment-16533248 ]
Yiqun Lin commented on HDFS-13674: ---------------------------------- Hi [~csun], I don't think it's a improvement to separate quantile metric from one to five. Since they are almost same, so we combine them in one line. It should be okay I think. {noformat} -| `EditLogFetchTime`*num*`s(50/75/90/95/99)thPercentileLatency` | The 50/75/90/95/99th percentile of time spent in fetching edit streams from journal nodes by standby NameNode, in milliseconds. Percentile measurement is off by default, by watching no intervals. The intervals are specified by `dfs.metrics.percentiles.intervals`. | +| `EditLogFetchTime`*num*`s50thPercentileLatency` | The 50th percentile of time spent in fetching edit streams from journal nodes by standby NameNode in milliseconds (*num* seconds granularity) if `rpc.metrics.quantile.enable` is set to true. *num* is specified by `rpc.metrics.percentiles.intervals`. | +| `EditLogFetchTime`*num*`s75thPercentileLatency` | The 75th percentile of time spent in fetching edit streams from journal nodes by standby NameNode in milliseconds (*num* seconds granularity) if `rpc.metrics.quantile.enable` is set to true. *num* is specified by `rpc.metrics.percentiles.intervals`. | +| `EditLogFetchTime`*num*`s90thPercentileLatency` | The 90th percentile of time spent in fetching edit streams from journal nodes by standby NameNode in milliseconds (*num* seconds granularity) if `rpc.metrics.quantile.enable` is set to true. *num* is specified by `rpc.metrics.percentiles.intervals`. | +| `EditLogFetchTime`*num*`s95thPercentileLatency` | The 95th percentile of time spent in fetching edit streams from journal nodes by standby NameNode in milliseconds (*num* seconds granularity) if `rpc.metrics.quantile.enable` is set to true. *num* is specified by `rpc.metrics.percentiles.intervals`. | +| `EditLogFetchTime`*num*`s99thPercentileLatency` | The 99th percentile of time spent in fetching edit streams from journal nodes by standby NameNode in milliseconds (*num* seconds granularity) if `rpc.metrics.quantile.enable` is set to true. *num* is specified by `rpc.metrics.percentiles.intervals`. | {noformat} Other change looks good to me. > Improve documentation on Metrics > -------------------------------- > > Key: HDFS-13674 > URL: https://issues.apache.org/jira/browse/HDFS-13674 > Project: Hadoop HDFS > Issue Type: Improvement > Components: documentation, metrics > Reporter: Chao Sun > Assignee: Chao Sun > Priority: Minor > Attachments: HDFS-13674.000.patch > > > There are a few confusing places in the [Hadoop Metrics > page|https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-common/Metrics.html]. > For instance, there are duplicated entries such as {{FsImageLoadTime}}; some > quantile metrics do not have corresponding entries, description on some > quantile metrics are not very specific on what is the {{num}} variable in the > metrics name, etc. > This JIRA targets at improving this. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org