GitHub user StephanEwen opened a pull request:
https://github.com/apache/flink/pull/5645
FLINK-8876 Improve concurrent access handling in stateful serializers
## What is the purpose of the change
Help detecting accidental concurrent use of serializers.
If the log is set to `DEBUG`, or assertions are enabled (as they are during
tests), the serializers will remember the current thread that performs
serialization/deserialization/copying and throw an error if another thread
accesses the serializer at the same time.
The code that implements that check is under a `static final` flag, so
should be eliminated by the JIT if the utility is not active, meaning it has
zero cost when not used. To initialize the flag, we use the logger (using the
fact that Java initializes static fields in order) and a utility class (using
the fact that the utility class is initialized before the serializer class.)
## Brief change log
- Add a concurrency detector to the KryoSerializer
- Add a concurrency detector to the AvroSerializer
## Verifying this change
The change adds unit tests to validate the effectiveness of the utility.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no)**
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no)**
- The serializers: **(yes** / no / don't know)
- The runtime per-record code paths (performance sensitive): **(yes** /
no / don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / **no** / don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? **(yes** / no)
- If yes, how is the feature documented? (**not applicable** / docs /
JavaDocs / not documented)
There is no change that users should make, this simply enhances test
coverage and results in more meaningful errors during debug runs.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/StephanEwen/incubator-flink kryo_safety
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5645.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 #5645
----
commit 4a99516ca938bf4cf6dfcdbd0e5e4615367c8eac
Author: Stephan Ewen <sewen@...>
Date: 2018-03-04T11:11:29Z
[FLINK-8877] [core] Set Kryo trace if Flink log level is TRACE
commit 3f20d2098de39997b89b79366a636c4dde77dc6a
Author: Stephan Ewen <sewen@...>
Date: 2018-03-06T10:30:54Z
[FLINK-8878] [tests] Add BlockerSync utility
This helps to synchronize two threads of which one is expected to block
while holding a resource.
commit 727003165447ee2a925b3030f19afe4e1b3b5325
Author: Stephan Ewen <sewen@...>
Date: 2018-03-04T11:20:17Z
[FLINK-8878] [core] Add concurrency check Kryo Serializer on DEBUG level
commit 714ee8882f86aa2a969dc842097dc11687d04607
Author: Stephan Ewen <sewen@...>
Date: 2018-03-06T10:21:08Z
[FLINK-8879] [avro] Add concurrency check Avro Serializer on DEBUG level.
----
---
