Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/1954#discussion_r67509184
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/executiongraph/restart/FailureRateRestartStrategy.java
---
@@ -0,0 +1,108 @@
+/*
+ * 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.executiongraph.restart;
+
+import com.google.common.base.Preconditions;
+import org.apache.flink.configuration.ConfigConstants;
+import org.apache.flink.configuration.Configuration;
+import org.apache.flink.runtime.executiongraph.ExecutionGraph;
+import scala.concurrent.duration.Duration;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.TimeUnit;
+
+import static akka.dispatch.Futures.future;
+
+/**
+ * Restart strategy which tries to restart the given {@link
ExecutionGraph} when failure rate exceeded
+ * with a fixed time delay in between.
+ */
+public class FailureRateRestartStrategy implements RestartStrategy {
+ private final int maxFailuresPerUnit;
+ private final TimeUnit failureRateUnit;
+ private final long delayBetweenRestartAttempts;
+ private List<Long> restartTimestamps = new ArrayList<>();
+ private boolean disabled = false;
+
+ public FailureRateRestartStrategy(int maxFailuresPerUnit, TimeUnit
failureRateUnit, long delayBetweenRestartAttempts) {
+ Preconditions.checkArgument(maxFailuresPerUnit > 0, "Maximum
number of restart attempts per time unit must be greater than 0.");
+ Preconditions.checkArgument(delayBetweenRestartAttempts >= 0,
"Delay between restart attempts must be positive");
+
+ this.maxFailuresPerUnit = maxFailuresPerUnit;
+ this.failureRateUnit = failureRateUnit;
+ this.delayBetweenRestartAttempts = delayBetweenRestartAttempts;
+ }
+
+ @Override
+ public boolean canRestart() {
+ return !disabled && canRestartJob();
+ }
+
+ private boolean canRestartJob() {
+ int restartsInWindowSoFar = restartTimestamps.size();
+ if (restartsInWindowSoFar >= maxFailuresPerUnit) {
+ List<Long> lastFailures =
restartTimestamps.subList(restartsInWindowSoFar - maxFailuresPerUnit,
restartsInWindowSoFar);
+ restartTimestamps = lastFailures; //deallocating not
needed timestamps
--- End diff --
This does not deallocate not needed timestamps, because `subList` only
returns a view on the `restartTimestamps` list. So internally, the original
list will always be referenced. Furthermore, each add might cause that the
whole list is copied to a new list which is one element larger.
I think we should use a fixed size queue (size = maxFailuresPerUnit).
Whenever we call `canRestartJob` we first pop all timestamps which fall out of
the current failure interval. If the resulting size of the queue is `<
maxFailuresPerUnit`, then we can restart. Otherwise we've seen more than
`maxFailurePerUnit` in the specified time interval.
---
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.
---
