[
https://issues.apache.org/jira/browse/FLINK-7689?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16497111#comment-16497111
]
ASF GitHub Bot commented on FLINK-7689:
---------------------------------------
Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/4725#discussion_r192222379
--- Diff:
flink-connectors/flink-jdbc/src/main/java/org/apache/flink/api/java/io/jdbc/JDBCOutputFormat.java
---
@@ -82,6 +97,22 @@ public void open(int taskNumber, int numTasks) throws
IOException {
} catch (ClassNotFoundException cnfe) {
throw new IllegalArgumentException("JDBC driver class
not found.", cnfe);
}
+ this.flushMeter = getRuntimeContext()
+ .getMetricGroup()
+ .addGroup(FLUSH_SCOPE)
+ .meter(FLUSH_RATE_METER_NAME, new
DropwizardMeterWrapper(new com.codahale.metrics.Meter()));
+ this.batchLimitReachedMeter = getRuntimeContext()
+ .getMetricGroup()
+ .addGroup(FLUSH_SCOPE)
+ .meter(BATCH_LIMIT_REACHED_RATE_METER_NAME, new
DropwizardMeterWrapper(new com.codahale.metrics.Meter()));
+ this.flushDurationMsHisto = getRuntimeContext()
+ .getMetricGroup()
+ .addGroup(FLUSH_SCOPE)
+ .histogram(FLUSH_DURATION_HISTO_NAME, new
DropwizardHistogramWrapper(new com.codahale.metrics.Histogram(new
ExponentiallyDecayingReservoir())));
--- End diff --
I recommend staying away form histograms as long as possible. Most metric
backends recommend to _not_ build histograms in the application, but let the
backend handle it.
> Instrument the Flink JDBC sink
> ------------------------------
>
> Key: FLINK-7689
> URL: https://issues.apache.org/jira/browse/FLINK-7689
> Project: Flink
> Issue Type: Improvement
> Components: Streaming Connectors
> Affects Versions: 1.4.0
> Reporter: Martin Eden
> Priority: Minor
> Labels: jdbc, metrics
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> As confirmed by the Flink community in the following mailing list
> [message|http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/metrics-for-Flink-sinks-td15200.html]
> using off the shelf Flink sinks like the JDBC sink, Redis sink or Cassandra
> sink etc does not expose any sink specific metrics.
> The purpose of this ticket is to add some relevant metrics to the
> JDBCOutputFormat:
> - Meters for when a flush is made.
> - Histograms for the jdbc batch count and batch execution latency.
> These would allow deeper understanding of the runtime behaviour of
> performance critical jobs writing to external databases using this generic
> interface.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
