You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by aljoscha <gi...@git.apache.org> on 2015/04/20 17:00:23 UTC
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
GitHub user aljoscha opened a pull request:
https://github.com/apache/flink/pull/612
[FLINK-1867/1880] Raise test timeouts in hope of fixing Travis fails
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/aljoscha/flink raise-test-timeouts
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/612.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #612
----
commit 4df27ee0ec1ce68376d51c4a882116651fd52788
Author: Aljoscha Krettek <al...@gmail.com>
Date: 2015-04-13T14:16:51Z
[FLINK-1867/1880] Raise test timeouts in hope of fixing Travis fails
----
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/612#issuecomment-94799019
Any thoughts on this? I would really like to merge this to improve Travis reliability.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943704
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/jobmanager/JobManagerFailsITCase.scala ---
@@ -18,24 +18,23 @@
package org.apache.flink.api.scala.runtime.jobmanager
-import akka.actor.Status.{Success, Failure}
+import akka.actor.Status.Success
import akka.actor.{ActorSystem, PoisonPill}
import akka.testkit.{ImplicitSender, TestKit}
+import org.junit.Ignore
+import org.junit.runner.RunWith
+import org.scalatest.junit.JUnitRunner
+import org.scalatest.{BeforeAndAfterAll, Matchers, WordSpecLike}
+
import org.apache.flink.configuration.{ConfigConstants, Configuration}
import org.apache.flink.runtime.akka.AkkaUtils
-import org.apache.flink.runtime.client.JobExecutionException
-import org.apache.flink.runtime.jobgraph.{JobGraph, AbstractJobVertex}
-import org.apache.flink.runtime.jobmanager.Tasks.{NoOpInvokable, BlockingNoOpInvokable}
+import org.apache.flink.runtime.jobgraph.{AbstractJobVertex, JobGraph}
+import org.apache.flink.runtime.jobmanager.Tasks.{BlockingNoOpInvokable, NoOpInvokable}
import org.apache.flink.runtime.messages.JobManagerMessages._
import org.apache.flink.runtime.testingUtils.TestingMessages.DisableDisconnect
-import org.apache.flink.runtime.testingUtils.TestingTaskManagerMessages.{JobManagerTerminated,
-NotifyWhenJobManagerTerminated}
+import org.apache.flink.runtime.testingUtils.TestingTaskManagerMessages.{JobManagerTerminated, NotifyWhenJobManagerTerminated}
import org.apache.flink.runtime.testingUtils.TestingUtils
import org.apache.flink.test.util.ForkableFlinkMiniCluster
-import org.junit.Ignore
-import org.junit.runner.RunWith
-import org.scalatest.junit.JUnitRunner
-import org.scalatest.{BeforeAndAfterAll, Matchers, WordSpecLike}
@Ignore("Contains a bug with Akka 2.2.1")
--- End diff --
True, removing it
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha closed the pull request at:
https://github.com/apache/flink/pull/612
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28814328
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/jobmanager/JobManagerFailsITCase.scala ---
@@ -18,24 +18,23 @@
package org.apache.flink.api.scala.runtime.jobmanager
-import akka.actor.Status.{Success, Failure}
+import akka.actor.Status.Success
import akka.actor.{ActorSystem, PoisonPill}
import akka.testkit.{ImplicitSender, TestKit}
+import org.junit.Ignore
+import org.junit.runner.RunWith
+import org.scalatest.junit.JUnitRunner
+import org.scalatest.{BeforeAndAfterAll, Matchers, WordSpecLike}
+
import org.apache.flink.configuration.{ConfigConstants, Configuration}
import org.apache.flink.runtime.akka.AkkaUtils
-import org.apache.flink.runtime.client.JobExecutionException
-import org.apache.flink.runtime.jobgraph.{JobGraph, AbstractJobVertex}
-import org.apache.flink.runtime.jobmanager.Tasks.{NoOpInvokable, BlockingNoOpInvokable}
+import org.apache.flink.runtime.jobgraph.{AbstractJobVertex, JobGraph}
+import org.apache.flink.runtime.jobmanager.Tasks.{BlockingNoOpInvokable, NoOpInvokable}
import org.apache.flink.runtime.messages.JobManagerMessages._
import org.apache.flink.runtime.testingUtils.TestingMessages.DisableDisconnect
-import org.apache.flink.runtime.testingUtils.TestingTaskManagerMessages.{JobManagerTerminated,
-NotifyWhenJobManagerTerminated}
+import org.apache.flink.runtime.testingUtils.TestingTaskManagerMessages.{JobManagerTerminated, NotifyWhenJobManagerTerminated}
import org.apache.flink.runtime.testingUtils.TestingUtils
import org.apache.flink.test.util.ForkableFlinkMiniCluster
-import org.junit.Ignore
-import org.junit.runner.RunWith
-import org.scalatest.junit.JUnitRunner
-import org.scalatest.{BeforeAndAfterAll, Matchers, WordSpecLike}
@Ignore("Contains a bug with Akka 2.2.1")
--- End diff --
Should this line not be removed by now?
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943824
--- Diff: flink-tests/src/test/java/org/apache/flink/test/recovery/AbstractProcessFailureRecoveryTest.java ---
@@ -112,9 +112,9 @@ public void testTaskManagerProcessFailure() {
Tuple2<String, Object> localAddress = new Tuple2<String, Object>("localhost", jobManagerPort);
Configuration jmConfig = new Configuration();
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 s");
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4 s");
- jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 2);
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 ms");
--- End diff --
That's a typo, will fix it.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/612#issuecomment-95946066
Manually merged.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943875
--- Diff: flink-tests/src/test/java/org/apache/flink/test/recovery/AbstractProcessFailureRecoveryTest.java ---
@@ -112,9 +112,9 @@ public void testTaskManagerProcessFailure() {
Tuple2<String, Object> localAddress = new Tuple2<String, Object>("localhost", jobManagerPort);
Configuration jmConfig = new Configuration();
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 s");
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4 s");
- jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 2);
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 ms");
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s");
+ jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 20);
--- End diff --
See above.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943598
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/taskmanager/TaskManagerFailsITCase.scala ---
@@ -231,9 +231,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
No, it doesn't before and after this change the 4 tests complete in about 8 seconds. @tillrohrmann suggested that, since the actor system is local, there are some other mechanisms in play that signal failure in this case.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943849
--- Diff: flink-tests/src/test/java/org/apache/flink/test/recovery/AbstractProcessFailureRecoveryTest.java ---
@@ -112,9 +112,9 @@ public void testTaskManagerProcessFailure() {
Tuple2<String, Object> localAddress = new Tuple2<String, Object>("localhost", jobManagerPort);
Configuration jmConfig = new Configuration();
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 s");
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4 s");
- jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 2);
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 ms");
--- End diff --
Seemed to run flawlessly, though. :smile: I can the tests about 100 times by now without seeing another failure in the targeted tests.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28961799
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/taskmanager/TaskManagerFailsITCase.scala ---
@@ -231,9 +231,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
Yes, removing them also from JobManagerFailsITCase
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28961064
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/taskmanager/TaskManagerFailsITCase.scala ---
@@ -231,9 +231,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
True, the heartbeats go only between actor systems, not actors. If everything is local, no hearbeats happen anyways.
May as well drop this config entry then, I suppose...
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by aljoscha <gi...@git.apache.org>.
Github user aljoscha commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28943731
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/jobmanager/JobManagerFailsITCase.scala ---
@@ -136,9 +135,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
As above, it doesn't affect test execution time.
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28814484
--- Diff: flink-tests/src/test/java/org/apache/flink/test/recovery/AbstractProcessFailureRecoveryTest.java ---
@@ -112,9 +112,9 @@ public void testTaskManagerProcessFailure() {
Tuple2<String, Object> localAddress = new Tuple2<String, Object>("localhost", jobManagerPort);
Configuration jmConfig = new Configuration();
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 s");
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4 s");
- jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 2);
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 ms");
--- End diff --
Is this really a good idea? Heartbeats every 1 ms? This is almost longer than the TCP process-to-process roundtrip time. May be a bit too aggressive...
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28814178
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/taskmanager/TaskManagerFailsITCase.scala ---
@@ -231,9 +231,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
What is the effect of this setting on the test? Does it take very long now, because the JobManager needs really long to determine that the TaskManager has failed?
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28814574
--- Diff: flink-tests/src/test/java/org/apache/flink/test/recovery/AbstractProcessFailureRecoveryTest.java ---
@@ -112,9 +112,9 @@ public void testTaskManagerProcessFailure() {
Tuple2<String, Object> localAddress = new Tuple2<String, Object>("localhost", jobManagerPort);
Configuration jmConfig = new Configuration();
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 s");
- jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4 s");
- jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 2);
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1 ms");
+ jmConfig.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s");
+ jmConfig.setInteger(ConfigConstants.AKKA_WATCH_THRESHOLD, 20);
--- End diff --
How long do the tests take now? With a pause of 20s and a threshold of 20, how long does the JobManager take to realize that the TaskManager is down?
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---
[GitHub] flink pull request: [FLINK-1867/1880] Raise test timeouts in hope ...
Posted by StephanEwen <gi...@git.apache.org>.
Github user StephanEwen commented on a diff in the pull request:
https://github.com/apache/flink/pull/612#discussion_r28814392
--- Diff: flink-tests/src/test/scala/org/apache/flink/api/scala/runtime/jobmanager/JobManagerFailsITCase.scala ---
@@ -136,9 +135,9 @@ with WordSpecLike with Matchers with BeforeAndAfterAll {
config.setInteger(ConfigConstants.TASK_MANAGER_NUM_TASK_SLOTS, numSlots)
config.setInteger(ConfigConstants.LOCAL_INSTANCE_MANAGER_NUMBER_TASK_MANAGER, numTaskmanagers)
config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_INTERVAL, "1000 ms")
- config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "4000 ms")
+ config.setString(ConfigConstants.AKKA_WATCH_HEARTBEAT_PAUSE, "20 s")
--- End diff --
What is the effect of this setting on the test? Does it take very long now, because the TaskManager needs really long to determine that the JobManager has failed?
---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---