flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-9004) Cluster test: Run general purpose job with failures with Yarn session
Date Mon, 02 Jul 2018 15:06:00 GMT

    [ https://issues.apache.org/jira/browse/FLINK-9004?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16530047#comment-16530047

ASF GitHub Bot commented on FLINK-9004:

Github user yanghua commented on a diff in the pull request:

    --- Diff: jepsen-flink/README.md ---
    @@ -0,0 +1,60 @@
    +# jepsen.flink
    +A Clojure project based on the [Jepsen](https://github.com/jepsen-io/jepsen) framework
to find bugs in the
    +distributed coordination of Apache Flink®.
    +## Test Coverage
    +Jepsen is a framework built to test the behavior of distributed systems
    +under faults. The tests in this particular project deploy Flink on either YARN or Mesos,
submit a
    +job, and examine the availability of the job after injecting faults.
    +A job is said to be available if all the tasks of the job are running.
    +The faults that can be currently introduced to the Flink cluster include:
    +* Killing of TaskManager/JobManager processes
    +* Stopping HDFS NameNode
    +* Network partitions
    +There are many more properties other than job availability that could be
    +verified but are not yet covered by this test suite, e.g., end-to-end exactly-once processing
    +## Usage
    +See the [Jepsen documentation](https://github.com/jepsen-io/jepsen#setting-up-a-jepsen-environment)
    +for how to set up the environment to run tests. The `scripts/run-tests.sh` documents
how to invoke
    +tests. The Flink job used for testing is located under
    +`flink-end-to-end-tests/flink-datastream-allround-test`. You have to build the job first
and copy
    +the resulting jar (`DataStreamAllroundTestProgram.jar`) to the `./bin` directory of this
    +To simplify development, we have prepared Dockerfiles and a Docker Compose template
    +so that you can run the tests locally in containers. To build the images
    +and start the containers, simply run:
    +    $ cd docker
    +    $ ./up.sh
    +After the containers started, open a new terminal window and run `docker exec -it jepsen-control
    +This will allow you to run arbitrary commands on the control node.
    +To start the tests, you can use the `run-tests.sh` script in the `docker` directory,
    +which expects the number of test iterations, and a URI to a Flink distribution, e.g.,
    +    ./docker/run-tests.sh 1 https://example.com/flink-dist.tgz
    +The project's root is mounted as a volume to all containers under the path `/jepsen`.
    +This means that changes to the test sources are immediately reflected in the control
node container.
    +Moreover, this allows you to test locally built Flink distributions by copying the tarball
to the
    +project's root and passing a URI with the `file://` scheme to the `run-tests.sh` script,
    +### Checking the output of tests
    +Consult the `jepsen.log` file for the particular test run in the `store` folder. The
final output of every test will be either
    +    Everything looks good! ヽ('ー`)ノ
    +    Analysis invalid! (ノಥ益ಥ)ノ ┻━┻
    --- End diff --
    the end of this line is gibberish?

> Cluster test: Run general purpose job with failures with Yarn session
> ---------------------------------------------------------------------
>                 Key: FLINK-9004
>                 URL: https://issues.apache.org/jira/browse/FLINK-9004
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Tests
>    Affects Versions: 1.5.0
>            Reporter: Till Rohrmann
>            Assignee: Gary Yao
>            Priority: Blocker
>              Labels: pull-request-available
>             Fix For: 1.6.0, 1.5.1
> Similar to FLINK-8973, we should run the general purpose job (FLINK-8971) on a Yarn session
cluster and simulate failures.
> The job jar should be ill-packaged, meaning that we include too many dependencies in
the user jar. We should include the Scala library, Hadoop and Flink itself to verify that
there are no class loading issues.
> The general purpose job should run with misbehavior activated. Additionally, we should
simulate at least the following failure scenarios:
> * Kill Flink processes
> * Kill connection to storage system for checkpoints and jobs
> * Simulate network partition
> We should run the test at least with the following state backend: RocksDB incremental
async and checkpointing to S3.

This message was sent by Atlassian JIRA

View raw message