Github user zentol commented on a diff in the pull request:
https://github.com/apache/flink/pull/3348#discussion_r105753418
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/consumer/InputGateMetrics.java
---
@@ -0,0 +1,168 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.runtime.io.network.partition.consumer;
+
+import org.apache.flink.metrics.Gauge;
+import org.apache.flink.metrics.MetricGroup;
+
+import static org.apache.flink.util.Preconditions.checkNotNull;
+
+public class InputGateMetrics {
+
+ private final SingleInputGate inputGate;
+
+ private long lastTotal = -1;
+
+ private int lastMin = -1;
+
+ private int lastMax = -1;
+
+ private float lastAvg = -1.0f;
+
+ //
------------------------------------------------------------------------
+
+ private InputGateMetrics(SingleInputGate inputGate) {
+ this.inputGate = checkNotNull(inputGate);
+ }
+
+ //
------------------------------------------------------------------------
+
+ // these methods are package private to make access from the nested
classes faster
+
+ long refreshAndGetTotal() {
+ long total;
+ if ((total = lastTotal) == -1) {
+ refresh();
--- End diff --
Custom objects can't be displayed properly in the web interface since we
call ```toString()``` on it. The same happens in most reporters; so this isn't
really an option.
As it stands we don't have a single metrics that is guaranteed to be 100%
consistent with other metrics. numRecordsOut and numBytesOut to not descriibe
the same moment in time. Neither is this guaranteed for the checkpoint metrics;
while these are updated all at once (from the outside), there is no mechanism
that prevents this update in the middle of a report.
I don't know a lot about the network stack; so whether it is truly
necessary to have all metrics describe one point in time I can't say.
If this is necessary the only way i can think of right now is abusing the
View metric type. View's are meant an add-on for metrics that want to be
updated in regular intervals (5 seconds) regardless of when their value is
actually requested. A metric that only implements the View interface is never
reported, but still updated, so you could have this view update a shared
data-structure from which the other gauges simply retrieve the current value,
If this is not necessary i would simply separate them and don't worry about
the performance overhead of the metrics; as long as this doesn't affect the job
via taking locks or similar.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
