GitHub user rmetzger opened a pull request:
https://github.com/apache/flink/pull/2065
[FLINK-4011] Keep UserCodeClassLoader in archived ExecutionGraphs
Currently, completed jobs cannot be accessed in the web frontend, because
the classloader passed to `SerializedValue` is always null.
There are different approaches to resolve this issue:
- Use the system classloader to deserialize the EC. This means that as soon
as the EC contains user code, we can not deserialize it. The web frontent will
show fewer information
- In `ExecutionGraph.prepareForArchiving()`, we deserialize the EC into a
special field, then we set the user code classloader free for GCing. This would
be a hacky approach because we would have two ECs (serialized, regular
instance) in the EG.
- Keep the usercodeclassloader in the EC. This means the classes of the job
can not be unloaded from the JobManager JVM until the job has been removed from
the JM history.
I'm open for discussing more approaches or alternative solutions.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/rmetzger/flink flink4011
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/2065.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 #2065
commit 388e75c95cd4e63419af429102e8afcfcd11cb8f
Author: Robert Metzger
Date: 2016-06-02T14:08:07Z
[FLINK-4011] Keep UserCodeClassLoader in archived ExecutionGraphs
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---