[jira] [Created] (FLINK-8182) Unable to read hdfs file system directory(which contains sub directories) recursively
Shashank Agarwal created FLINK-8182: --- Summary: Unable to read hdfs file system directory(which contains sub directories) recursively Key: FLINK-8182 URL: https://issues.apache.org/jira/browse/FLINK-8182 Project: Flink Issue Type: Improvement Components: Streaming Affects Versions: 1.3.2 Reporter: Shashank Agarwal Unable to read hdfs file system directory(which contains subdirectories) recursively, It works fine when a single directory contains only files but when the directory contains subdirectories it dosesn't read subdirectory files. {code} streamExecutionEnvironment.readTextFile("HDFS path") {code} -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-8183) Add native Avro type support to the Table API & SQL
Timo Walther created FLINK-8183: --- Summary: Add native Avro type support to the Table API & SQL Key: FLINK-8183 URL: https://issues.apache.org/jira/browse/FLINK-8183 Project: Flink Issue Type: New Feature Components: Table API & SQL Reporter: Timo Walther Avro types can pass the Table API, however, there should be a more native support in order to have the best user experience. This issue is an umbrella issue for tasks that would improve the handling of Avro types: Improvements could be: - Create a Avro type information that is created from an Avro schema that maps to all supported Table API types (full knowledge about key and values of lists, maps, and union types instead of {{GenericType}}) - Convert {{Utf8}} (even in nested Avro types) to string when entering the Table API and convert it back if necessary - Add scalar functions to change certain values (e.g., {{select('avroRecord.set("name", "Bob").set("age", 12))}}). This is in particular useful when a type has a lot of fields. -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-8184) Return raw JsonPlan instead of escaped string value in JobDetailsInfo
Till Rohrmann created FLINK-8184: Summary: Return raw JsonPlan instead of escaped string value in JobDetailsInfo Key: FLINK-8184 URL: https://issues.apache.org/jira/browse/FLINK-8184 Project: Flink Issue Type: Bug Components: REST Affects Versions: 1.5.0 Reporter: Till Rohrmann Assignee: Till Rohrmann Priority: Minor Fix For: 1.5.0 The {{JobDetailsInfo}} should pass the JsonPlan as a raw value because otherwise the string value will be escaped. -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-8185) Make first level flattening optional for input and output
Timo Walther created FLINK-8185: --- Summary: Make first level flattening optional for input and output Key: FLINK-8185 URL: https://issues.apache.org/jira/browse/FLINK-8185 Project: Flink Issue Type: Improvement Components: Table API & SQL Reporter: Timo Walther When converting a {{DataSet}}/{{DataStream}} into a table. A composite type is automatically flattened to a row of fields, e.g. {{MyPojo}} becomes {{Row}}. There is no possibility to keep {{Row>}}. This would be especially interesting for POJOs that should not be flattened and be converted back to the same type: {{toTable[MyPojo]}}. At the moment a user has to specify all fields in order to get the same behavior. -- This message was sent by Atlassian JIRA (v6.4.14#64029)
Flink 1.4.0-RC2 Hadoop Build
Hello, I am trying to build and run Flink from 1.4.0-rc2 branch with hadoop binary 2.7.0 compatibility. Here are the steps I followed to build (I have maven 3.3.9). ===cd $FLINK_HOMEmvn clean install -DskipTests -Dhadoop.version=2.7.0cd $FLINK_HOME/flink-distmvn clean install -Dhadoop.version=2.7.0 Running Flink from $FLINK_HOME/flink-dist/target/flink-1.4.0-bin== I am seeing below error error messages from the logs and it suggests something to do with Hadoop dependency not available? Could someone please confirm the build steps? org.apache.flink.core.fs.UnsupportedFileSystemSchemeException: Could not find a file system implementation for scheme 'hdfs'. The scheme is not directly supported by Flink and no Hadoop file system to support this scheme could be loaded. at org.apache.flink.core.fs.FileSystem.getUnguardedFileSystem(FileSystem.java:403) at org.apache.flink.core.fs.FileSystem.get(FileSystem.java:318) at org.apache.flink.core.fs.Path.getFileSystem(Path.java:293) at org.apache.flink.runtime.state.filesystem.FsCheckpointStreamFactory.(FsCheckpointStreamFactory.java:99) at org.apache.flink.runtime.state.filesystem.FsStateBackend.createStreamFactory(FsStateBackend.java:277) at org.apache.flink.streaming.runtime.tasks.StreamTask.createCheckpointStreamFactory(StreamTask.java:787) at org.apache.flink.streaming.api.operators.AbstractStreamOperator.initializeState(AbstractStreamOperator.java:247) at org.apache.flink.streaming.runtime.tasks.StreamTask.initializeOperators(StreamTask.java:694) at org.apache.flink.streaming.runtime.tasks.StreamTask.initializeState(StreamTask.java:682) at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:253) at org.apache.flink.runtime.taskmanager.Task.run(Task.java:718) at java.lang.Thread.run(Thread.java:748)Caused by: org.apache.flink.core.fs.UnsupportedFileSystemSchemeException: Cannot support file system for 'hdfs' via Hadoop, because Hadoop is not in the classpath, or some classes are missing from the classpath. at org.apache.flink.runtime.fs.hdfs.HadoopFsFactory.create(HadoopFsFactory.java:179) at org.apache.flink.core.fs.FileSystem.getUnguardedFileSystem(FileSystem.java:399) ... 11 moreCaused by: java.lang.NoClassDefFoundError: Could not initialize class org.apache.hadoop.hdfs.DFSConfigKeys at org.apache.hadoop.hdfs.DFSClient$Conf.(DFSClient.java:509) at org.apache.hadoop.hdfs.DFSClient.(DFSClient.java:637) at org.apache.hadoop.hdfs.DFSClient.(DFSClient.java:619) at org.apache.hadoop.hdfs.DistributedFileSystem.initialize(DistributedFileSystem.java:149) at org.apache.flink.runtime.fs.hdfs.HadoopFsFactory.create(HadoopFsFactory.java:159) ... 12 more2017-12-01 04:28:09,274 INFO org.apache.flink.runtime.executiongraph.ExecutionGraph - Source: Collection Source -> Map (1/1) (74e8a3f21c86e8dec3c55988fff42e5d) switched from RUNNING to FAILED.org.apache.flink.core.fs.UnsupportedFileSystemSchemeException: Could not find a file system implementation for scheme 'hdfs'. The scheme is not directly supported by Flink and no Hadoop file system to support this scheme could be loaded. at org.apache.flink.core.fs.FileSystem.getUnguardedFileSystem(FileSystem.java:403) at org.apache.flink.core.fs.FileSystem.get(FileSystem.java:318) at org.apache.flink.core.fs.Path.getFileSystem(Path.java:293) at org.apache.flink.runtime.state.filesystem.FsCheckpointStreamFactory.(FsCheckpointStreamFactory.java:99) at org.apache.flink.runtime.state.filesystem.FsStateBackend.createStreamFactory(FsStateBackend.java:277) at org.apache.flink.streaming.runtime.tasks.StreamTask.createCheckpointStreamFactory(StreamTask.java:787) at org.apache.flink.streaming.api.operators.AbstractStreamOperator.initializeState(AbstractStreamOperator.java:247) at org.apache.flink.streaming.runtime.tasks.StreamTask.initializeOperators(StreamTask.java:694) at org.apache.flink.streaming.runtime.tasks.StreamTask.initializeState(StreamTask.java:682) at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:253) at org.apache.flink.runtime.taskmanager.Task.run(Task.java:718) at java.lang.Thread.run(Thread.java:748)Caused by: org.apache.flink.core.fs.UnsupportedFileSystemSchemeException: Cannot support file system for 'hdfs' via Hadoop, because Hadoop is not in the classpath, or some classes are missing from the classpath. at org.apache.flink.runtime.fs.hdfs.HadoopFsFactory.create(HadoopFsFactory.java:179) at org.apache.flink.core.fs.FileSystem.getUnguardedFileSystem(FileSystem.java:399) ... 11 moreCaused by: java.lang.NoClassDefFoundError: Could not initialize class org.apache.hadoop.hdfs.DFSConfigKeys at org.apache.hadoop.hdfs.DFSClient$Conf.(DFSClient.java:509) at org.apache.hadoop.hdfs.DFSClient.(DFSClient.java:637) at org.apache.hadoop.hdfs.DFSClient.(DFSClient.java:
[jira] [Created] (FLINK-8186) AvroInputFormat regression: fails to deserialize GenericRecords on cluster with hadoop27 compat
Sebastian Klemke created FLINK-8186: --- Summary: AvroInputFormat regression: fails to deserialize GenericRecords on cluster with hadoop27 compat Key: FLINK-8186 URL: https://issues.apache.org/jira/browse/FLINK-8186 Project: Flink Issue Type: Bug Affects Versions: 1.4.0 Reporter: Sebastian Klemke The following job runs fine on a Flink 1.3.2 cluster, but fails on a Flink 1.4.0 RC2 standalone cluster, "hadoop27" flavour: {code} public class GenericRecordCount { public static void main(String[] args) throws Exception { String input = ParameterTool.fromArgs(args).getRequired("input"); ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment(); long count = env.readFile(new AvroInputFormat<>(new Path(input), GenericRecord.class), input) .count(); System.out.printf("Counted %d records\n", count); } } {code} Runs fine in LocalExecutionEnvironment and also on no-hadoop flavour standalone cluster, though. Exception thrown in Flink 1.4.0 hadoop27: {code} 12/01/2017 13:22:09 DataSource (at readFile(ExecutionEnvironment.java:514) (org.apache.flink.formats.avro.AvroInputFormat))(4/4) switched to FAILED java.lang.RuntimeException: java.lang.NoSuchMethodException: org.apache.avro.generic.GenericRecord.() at org.apache.avro.specific.SpecificData.newInstance(SpecificData.java:353) at org.apache.avro.specific.SpecificData.newRecord(SpecificData.java:369) at org.apache.avro.reflect.ReflectData.newRecord(ReflectData.java:901) at org.apache.avro.generic.GenericDatumReader.readRecord(GenericDatumReader.java:212) at org.apache.avro.generic.GenericDatumReader.readWithoutConversion(GenericDatumReader.java:175) at org.apache.avro.generic.GenericDatumReader.read(GenericDatumReader.java:153) at org.apache.avro.generic.GenericDatumReader.read(GenericDatumReader.java:145) at org.apache.avro.file.DataFileStream.next(DataFileStream.java:233) at org.apache.flink.formats.avro.AvroInputFormat.nextRecord(AvroInputFormat.java:165) at org.apache.flink.runtime.operators.DataSourceTask.invoke(DataSourceTask.java:167) at org.apache.flink.runtime.taskmanager.Task.run(Task.java:718) at java.lang.Thread.run(Thread.java:748) Caused by: java.lang.NoSuchMethodException: org.apache.avro.generic.GenericRecord.() at java.lang.Class.getConstructor0(Class.java:3082) at java.lang.Class.getDeclaredConstructor(Class.java:2178) at org.apache.avro.specific.SpecificData.newInstance(SpecificData.java:347) ... 11 more {code} -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-8187) Web client does not print errors
Timo Walther created FLINK-8187: --- Summary: Web client does not print errors Key: FLINK-8187 URL: https://issues.apache.org/jira/browse/FLINK-8187 Project: Flink Issue Type: Bug Components: Webfrontend Reporter: Timo Walther When submitting a jar with no defined Main class, the web client does not respond anymore and instead of printing the REST error: {code} java.util.concurrent.CompletionException: org.apache.flink.util.FlinkException: Could not run the jar. at org.apache.flink.runtime.webmonitor.handlers.JarRunHandler.lambda$handleJsonRequest$0(JarRunHandler.java:90) at java.util.concurrent.CompletableFuture$AsyncSupply.run(CompletableFuture.java:1590) at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180) at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) Caused by: org.apache.flink.util.FlinkException: Could not run the jar. ... 9 more Caused by: org.apache.flink.client.program.ProgramInvocationException: Neither a 'Main-Class', nor a 'program-class' entry was found in the jar file. at org.apache.flink.client.program.PackagedProgram.getEntryPointClassNameFromJar(PackagedProgram.java:592) at org.apache.flink.client.program.PackagedProgram.(PackagedProgram.java:188) at org.apache.flink.client.program.PackagedProgram.(PackagedProgram.java:147) at org.apache.flink.runtime.webmonitor.handlers.JarActionHandler.getJobGraphAndClassLoader(JarActionHandler.java:72) at org.apache.flink.runtime.webmonitor.handlers.JarRunHandler.lambda$handleJsonRequest$0(JarRunHandler.java:69) ... 8 more {code} -- This message was sent by Atlassian JIRA (v6.4.14#64029)
Re: [VOTE] Release 1.4.0, release candidate #2
Update on reported Mesos issue (FLINK-8174): TLDR; a PR will be ready within 24 hours that will undo reservation support. A couple of months ago, a fix (FLINK-7294) was merged related to how Flink accepts Mesos resource offers. The intention was to allow Flink to make use of so-called +reserved+ resources, a Mesos feature which makes it possible to reserve hosts for use by a specific framework/role. The fix inadvertently regressed the ability to use +unreserved+ resources. This is a serious regression because unreserved resources are the common case. The simple solution is to revert the earlier fix, deferring support for reservations to another release. We are spending some time to find a fix that works for all scenarios, but seems unlikely at this time. I am reaching out to the original contributor to get their feedback. In the course of the investigation, a related flaw was discovered in Fenzo that causes Flink to misinterpret offers that contain a mix of reserved and unreserved resources. I believe that a small fix is possible purely within Flink; an update to Fenzo does not appear necessary. Going forward, we will contribute an improved integration test suite with which to test Flink under diverse Mesos conditions (e.g. reservations). Thanks, Eron On Thu, Nov 30, 2017 at 9:47 PM, Tzu-Li (Gordon) Tai wrote: > Hi, > > I’ve noticed a behavioral regression in the Kafka producer, that should > also be considered a blocker: https://issues.apache.org/ > jira/browse/FLINK-8181 > There’s already a PR for the issue here: https://github.com/ > apache/flink/pull/5108 > > Best, > Gordon > > On 30 November 2017 at 5:27:22 PM, Fabian Hueske (fhue...@gmail.com) > wrote: > > I've created a JIRA issue for the the Hadoop 2.9.0 build problem [1]. > > Best, Fabian > > [1] https://issues.apache.org/jira/browse/FLINK-8177 > > 2017-11-30 4:35 GMT+01:00 Eron Wright : > > > Unfortunately we've identified a blocker bug for Flink on Mesos - > > FLINK-8174. We'll have a patch ready on Thursday. > > > > Thanks, > > Eron > > > > On Wed, Nov 29, 2017 at 3:40 PM, Eron Wright > wrote: > > > > > On Dell EMC side, we're testing the RC2 on DCOS 1.10.0. Seeing a > > > potential issue with offer acceptance and we'll update the thread with > a > > +1 > > > or with a more concrete issue within 24 hours. > > > > > > Thanks, > > > Eron > > > > > > On Wed, Nov 29, 2017 at 6:54 AM, Chesnay Schepler > > > wrote: > > > > > >> I don't think anyone has taken a look yet, nor was there a discussion > as > > >> to postponing it. > > >> > > >> It just slipped through the cracks i guess... > > >> > > >> > > >> On 29.11.2017 15:47, Gyula Fóra wrote: > > >> > > >>> Hi guys, > > >>> I ran into this again while playing with savepoint/restore > parallelism: > > >>> > > >>> https://issues.apache.org/jira/browse/FLINK-7595 > > >>> https://github.com/apache/flink/pull/4651 > > >>> > > >>> Anyone has some idea about the status of this PR or were we planning > to > > >>> postpone this to 1.5? > > >>> > > >>> Thanks, > > >>> Gyula > > >>> > > >>> > > >>> Fabian Hueske ezt írta (időpont: 2017. nov. 29., > > >>> Sze, > > >>> 13:10): > > >>> > > >>> OK, the situation is the following: > > > > The test class (org.apache.flink.yarn.UtilsTest) implements a > Hadoop > > interface (Container) that was extended in Hadoop 2.9.0 by a getter > > and > > setter. > > By adding the methods, we can compile Flink for Hadoop 2.9.0. > However, > > the > > getter/setter add a dependency on a class that was also added in > > Hadoop > > 2.9.0. > > Therefore, the implementation is not backwards compatible with > Hadoop > > versions < 2.9.0. > > > > Not sure how we can fix the problem. We would need two version of > the > > class > > that are chosen based on the Hadoop version. Do we have something > like > > that > > somewhere else? > > > > Since this is only a problem in a test class, Flink 1.4.0 might > still > > work > > very well with Hadoop 2.9.0. > > However, this has not been tested AFAIK. > > > > Cheers, Fabian > > > > 2017-11-29 12:47 GMT+01:00 Fabian Hueske : > > > > I just tried to build the release-1.4 branch for Hadoop 2.9.0 > > (released > > > a > > > few days ago) and got a compilation failure in a test class. > > > > > > Right now, I'm assessing how much we need to fix to support Hadoop > > > 2.9.0. > > > I'll report later. > > > > > > Best, Fabian > > > > > > 2017-11-29 11:16 GMT+01:00 Aljoscha Krettek : > > > > > > Agreed, this is a regression compared to the previous > functionality. > > I > > >> updated the issue to "Blocker". > > >> > > >> On 29. Nov 2017, at 10:01, Gyula Fóra > wrote: > > >>> > > >>> Hi all, > > >>> > > >>> I have found the following issue: > > >>> https://issues.apache.org/jira/browse/FLINK-8165 > > >>> > > >
[jira] [Created] (FLINK-8188) Clean up flink-contrib
Bowen Li created FLINK-8188: --- Summary: Clean up flink-contrib Key: FLINK-8188 URL: https://issues.apache.org/jira/browse/FLINK-8188 Project: Flink Issue Type: Improvement Affects Versions: 1.5.0 Reporter: Bowen Li This is the umbrella ticket for cleaning up flink-contrib. We argue that flink-contrib should be removed and all its submodules should be migrated to other top-level modules for the following reasons: 1) Apache Flink the whole project itself is a result of contributions from many developers, there's no reason to highlight some contributions in a dedicated module named 'contrib' 2) flink-contrib is already too crowded and noisy. It contains lots of sub modules with different purposes which confuse developers and users, and they lack a proper project hierarchy 3) This will save us quite some build time More details in discussions at FLINK-8175 and FLINK-8167 -- This message was sent by Atlassian JIRA (v6.4.14#64029)
[jira] [Created] (FLINK-8189) move flink-statebackend-rocksdb out of flink-contrib
Bowen Li created FLINK-8189: --- Summary: move flink-statebackend-rocksdb out of flink-contrib Key: FLINK-8189 URL: https://issues.apache.org/jira/browse/FLINK-8189 Project: Flink Issue Type: Sub-task Components: State Backends, Checkpointing Affects Versions: 1.5.0 Reporter: Bowen Li Assignee: Bowen Li Move {{flink-statebackend-rocksdb}} into probably its own/state-backend module or {{flink-runtime}} package. -- This message was sent by Atlassian JIRA (v6.4.14#64029)
Re: [VOTE] Release 1.4.0, release candidate #2
Thanks for the update! Just to be clear, you're proposing going forward with the "simple fix" of reverting FLINK-7294? > On 1. Dec 2017, at 18:39, Eron Wright wrote: > > Update on reported Mesos issue (FLINK-8174): > > TLDR; a PR will be ready within 24 hours that will undo reservation support. > > A couple of months ago, a fix (FLINK-7294) was merged related to how Flink > accepts Mesos resource offers. The intention was to allow Flink to make > use of so-called +reserved+ resources, a Mesos feature which makes it > possible to reserve hosts for use by a specific framework/role. The fix > inadvertently regressed the ability to use +unreserved+ resources. This is > a serious regression because unreserved resources are the common case. > > The simple solution is to revert the earlier fix, deferring support for > reservations to another release. We are spending some time to find a fix > that works for all scenarios, but seems unlikely at this time. I am > reaching out to the original contributor to get their feedback. > > In the course of the investigation, a related flaw was discovered in Fenzo > that causes Flink to misinterpret offers that contain a mix of reserved and > unreserved resources. I believe that a small fix is possible purely > within Flink; an update to Fenzo does not appear necessary. > > Going forward, we will contribute an improved integration test suite with > which to test Flink under diverse Mesos conditions (e.g. reservations). > > Thanks, > Eron > > On Thu, Nov 30, 2017 at 9:47 PM, Tzu-Li (Gordon) Tai > wrote: > >> Hi, >> >> I’ve noticed a behavioral regression in the Kafka producer, that should >> also be considered a blocker: https://issues.apache.org/ >> jira/browse/FLINK-8181 >> There’s already a PR for the issue here: https://github.com/ >> apache/flink/pull/5108 >> >> Best, >> Gordon >> >> On 30 November 2017 at 5:27:22 PM, Fabian Hueske (fhue...@gmail.com) >> wrote: >> >> I've created a JIRA issue for the the Hadoop 2.9.0 build problem [1]. >> >> Best, Fabian >> >> [1] https://issues.apache.org/jira/browse/FLINK-8177 >> >> 2017-11-30 4:35 GMT+01:00 Eron Wright : >> >>> Unfortunately we've identified a blocker bug for Flink on Mesos - >>> FLINK-8174. We'll have a patch ready on Thursday. >>> >>> Thanks, >>> Eron >>> >>> On Wed, Nov 29, 2017 at 3:40 PM, Eron Wright >> wrote: >>> On Dell EMC side, we're testing the RC2 on DCOS 1.10.0. Seeing a potential issue with offer acceptance and we'll update the thread with >> a >>> +1 or with a more concrete issue within 24 hours. Thanks, Eron On Wed, Nov 29, 2017 at 6:54 AM, Chesnay Schepler wrote: > I don't think anyone has taken a look yet, nor was there a discussion >> as > to postponing it. > > It just slipped through the cracks i guess... > > > On 29.11.2017 15:47, Gyula Fóra wrote: > >> Hi guys, >> I ran into this again while playing with savepoint/restore >> parallelism: >> >> https://issues.apache.org/jira/browse/FLINK-7595 >> https://github.com/apache/flink/pull/4651 >> >> Anyone has some idea about the status of this PR or were we planning >> to >> postpone this to 1.5? >> >> Thanks, >> Gyula >> >> >> Fabian Hueske ezt írta (időpont: 2017. nov. 29., >> Sze, >> 13:10): >> >> OK, the situation is the following: >>> >>> The test class (org.apache.flink.yarn.UtilsTest) implements a >> Hadoop >>> interface (Container) that was extended in Hadoop 2.9.0 by a getter >>> and >>> setter. >>> By adding the methods, we can compile Flink for Hadoop 2.9.0. >> However, >>> the >>> getter/setter add a dependency on a class that was also added in >>> Hadoop >>> 2.9.0. >>> Therefore, the implementation is not backwards compatible with >> Hadoop >>> versions < 2.9.0. >>> >>> Not sure how we can fix the problem. We would need two version of >> the >>> class >>> that are chosen based on the Hadoop version. Do we have something >> like >>> that >>> somewhere else? >>> >>> Since this is only a problem in a test class, Flink 1.4.0 might >> still >>> work >>> very well with Hadoop 2.9.0. >>> However, this has not been tested AFAIK. >>> >>> Cheers, Fabian >>> >>> 2017-11-29 12:47 GMT+01:00 Fabian Hueske : >>> >>> I just tried to build the release-1.4 branch for Hadoop 2.9.0 >>> (released a few days ago) and got a compilation failure in a test class. Right now, I'm assessing how much we need to fix to support Hadoop 2.9.0. I'll report later. Best, Fabian 2017-11-29 11:16 GMT+01:00 Aljoscha Krettek : Agreed, this is a regression compared to the previous >> functionality. >>> I > updated the issue to "Blocker". > >>
Re: [VOTE] Release 1.4.0, release candidate #2
There are three levels of support we could land on. 1. Flink works with unreserved resources (revert FLINK-7294). 2. Flink works with unreserved resources, and correctly ignores reserved resources (revert FLINK-7294 and mitigate Fenzo bug). 3. Flink works with unreserved resources and reserved resources. 3 is a moon shot. Striving for 2. Fallback on 1. On Fri, Dec 1, 2017 at 2:10 PM, Aljoscha Krettek wrote: > Thanks for the update! > > Just to be clear, you're proposing going forward with the "simple fix" of > reverting FLINK-7294? > > > On 1. Dec 2017, at 18:39, Eron Wright wrote: > > > > Update on reported Mesos issue (FLINK-8174): > > > > TLDR; a PR will be ready within 24 hours that will undo reservation > support. > > > > A couple of months ago, a fix (FLINK-7294) was merged related to how > Flink > > accepts Mesos resource offers. The intention was to allow Flink to make > > use of so-called +reserved+ resources, a Mesos feature which makes it > > possible to reserve hosts for use by a specific framework/role. The fix > > inadvertently regressed the ability to use +unreserved+ resources. This > is > > a serious regression because unreserved resources are the common case. > > > > The simple solution is to revert the earlier fix, deferring support for > > reservations to another release. We are spending some time to find a > fix > > that works for all scenarios, but seems unlikely at this time. I am > > reaching out to the original contributor to get their feedback. > > > > In the course of the investigation, a related flaw was discovered in > Fenzo > > that causes Flink to misinterpret offers that contain a mix of reserved > and > > unreserved resources. I believe that a small fix is possible purely > > within Flink; an update to Fenzo does not appear necessary. > > > > Going forward, we will contribute an improved integration test suite with > > which to test Flink under diverse Mesos conditions (e.g. reservations). > > > > Thanks, > > Eron > > > > On Thu, Nov 30, 2017 at 9:47 PM, Tzu-Li (Gordon) Tai < > tzuli...@apache.org> > > wrote: > > > >> Hi, > >> > >> I’ve noticed a behavioral regression in the Kafka producer, that should > >> also be considered a blocker: https://issues.apache.org/ > >> jira/browse/FLINK-8181 > >> There’s already a PR for the issue here: https://github.com/ > >> apache/flink/pull/5108 > >> > >> Best, > >> Gordon > >> > >> On 30 November 2017 at 5:27:22 PM, Fabian Hueske (fhue...@gmail.com) > >> wrote: > >> > >> I've created a JIRA issue for the the Hadoop 2.9.0 build problem [1]. > >> > >> Best, Fabian > >> > >> [1] https://issues.apache.org/jira/browse/FLINK-8177 > >> > >> 2017-11-30 4:35 GMT+01:00 Eron Wright : > >> > >>> Unfortunately we've identified a blocker bug for Flink on Mesos - > >>> FLINK-8174. We'll have a patch ready on Thursday. > >>> > >>> Thanks, > >>> Eron > >>> > >>> On Wed, Nov 29, 2017 at 3:40 PM, Eron Wright > >> wrote: > >>> > On Dell EMC side, we're testing the RC2 on DCOS 1.10.0. Seeing a > potential issue with offer acceptance and we'll update the thread with > >> a > >>> +1 > or with a more concrete issue within 24 hours. > > Thanks, > Eron > > On Wed, Nov 29, 2017 at 6:54 AM, Chesnay Schepler > > wrote: > > > I don't think anyone has taken a look yet, nor was there a discussion > >> as > > to postponing it. > > > > It just slipped through the cracks i guess... > > > > > > On 29.11.2017 15:47, Gyula Fóra wrote: > > > >> Hi guys, > >> I ran into this again while playing with savepoint/restore > >> parallelism: > >> > >> https://issues.apache.org/jira/browse/FLINK-7595 > >> https://github.com/apache/flink/pull/4651 > >> > >> Anyone has some idea about the status of this PR or were we planning > >> to > >> postpone this to 1.5? > >> > >> Thanks, > >> Gyula > >> > >> > >> Fabian Hueske ezt írta (időpont: 2017. nov. > 29., > >> Sze, > >> 13:10): > >> > >> OK, the situation is the following: > >>> > >>> The test class (org.apache.flink.yarn.UtilsTest) implements a > >> Hadoop > >>> interface (Container) that was extended in Hadoop 2.9.0 by a getter > >>> and > >>> setter. > >>> By adding the methods, we can compile Flink for Hadoop 2.9.0. > >> However, > >>> the > >>> getter/setter add a dependency on a class that was also added in > >>> Hadoop > >>> 2.9.0. > >>> Therefore, the implementation is not backwards compatible with > >> Hadoop > >>> versions < 2.9.0. > >>> > >>> Not sure how we can fix the problem. We would need two version of > >> the > >>> class > >>> that are chosen based on the Hadoop version. Do we have something > >> like > >>> that > >>> somewhere else? > >>> > >>> Since this is only a problem in a test class, Flink 1.4.0 might > >> still > >>> work > >>> very well with Hadoop 2.9.0. > >>