[ https://issues.apache.org/jira/browse/FLINK-12680?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16860210#comment-16860210 ]
Piyush Goyal commented on FLINK-12680: -------------------------------------- [~till.rohrmann] A good way to look at it is micrometer (which provides a standard facade over various different clients like prometheus, graphite, atlas etc.) and is used to abstract the specific implementation. it's like SLF4j for metrics. https://micrometer.io/docs/concepts#_timers The timer captures richer information which is not possible to be acquired using counter. All implementations of {{Timer}} report at least the total time and count of events as separate time series. > Add timer type to metric system > ------------------------------- > > Key: FLINK-12680 > URL: https://issues.apache.org/jira/browse/FLINK-12680 > Project: Flink > Issue Type: Improvement > Components: Runtime / Metrics > Reporter: Piyush Goyal > Priority: Major > > Currently Flink metric system supports counter, gauge, histogram and meter. > For cases which involve measuring the latency of short-term events (such as > request latency in the context of a connector), none of these metric types is > a good fit. Most metric system supports a timer type which is highly useful > in such cases. Timers are useful for measuring short-duration latencies and > the frequency of such events. All implementations of {{Timer}} report at > least the total time and count of events as separate time series. > Timer type will be a nice addition to the existing metric types in Flink. We > are happy to contribute it if there is agreement on it being a good addition. -- This message was sent by Atlassian JIRA (v7.6.3#76005)