[
https://issues.apache.org/jira/browse/FLINK-32649?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
chenyuzhi updated FLINK-32649:
------------------------------
Component/s: Runtime / Metrics
> Mismatch label and value for prometheus reporter
> ------------------------------------------------
>
> Key: FLINK-32649
> URL: https://issues.apache.org/jira/browse/FLINK-32649
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Metrics
> Affects Versions: 1.17.0
> Reporter: chenyuzhi
> Priority: Major
>
> When runing unit test
> 'org.apache.flink.metrics.prometheus.PrometheusReporterTest#metricIsRemovedWhileOtherMetricsWithSameNameExist',
> it got wrong response string as
> {code:java}
> # HELP flink_logical_scope_metric metric (scope: logical_scope)
> # TYPE flink_logical_scope_metric gauge
> flink_logical_scope_metric{label1="some_value",label2="value1",} 0.0 {code}
>
> in my opinion, the expected right response is :
>
> {code:java}
> # HELP flink_logical_scope_metric metric (scope: logical_scope)
> # TYPE flink_logical_scope_metric gauge
> flink_logical_scope_metric{label1="value1",label2="some_value",} 0.0
> {code}
>
>
> The reason may be we create two metric with same name, but two different
> order label-keys in the variables of MetricGroup. And we don't sort the
> key&value in variables within methond
> 'org.apache.flink.metrics.prometheus.AbstractPrometheusReporter#notifyOfAddedMetric'.
>
> Maybe it won't happen in production env now, cause we alway use the same
> 'MetricGroup' instance as the input param of method 'notifyOfAddedMetric'.
> Howerver it's important to ensure the robustness of method with unpected
> input param.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
