Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[SPARK-18959] the new leader will lost the statistics of the driver's resource on the worker When the leader master has changed. #16525

Closed
wants to merge 1 commit into from

Conversation

hustfxj
Copy link
Contributor

@hustfxj hustfxj commented Jan 10, 2017

I deploy the standalone cluster with two masters. and utilize zooKeeper to provide leader election. Firstly, I submit the application with cluster mode. Then I kill the leader master, and the standby master will be the leader. But the new leader will lost the statistics of the driver's resource. Then I stop the application, we will see the negative used resource at the worker from masterPage. Like that:

Workers

Worker Id	Address	State	Cores	Memory
worker-20161220162751-10.125.6.222-59295	10.125.6.222:59295	ALIVE	4 (-1 Used)	6.8 GB (-1073741824.0 B Used)
worker-20161220164233-10.218.135.80-10944	10.218.135.80:10944	ALIVE	4 (0 Used)	6.8 GB (0.0 B Used)

Because the new leader forget calculate the driver‘ resource when the master receive the "WorkerLatestState" message. At the same time we can set RUNNING state for the app after the master receive the message, otherwise the app' state will still be WAITTING.

…the worker When the leader master has changed.
@hustfxj hustfxj changed the title the new leader will lost the statistics of the driver's resource on the worker When the leader master has changed. [SPARK-18959] the new leader will lost the statistics of the driver's resource on the worker When the leader master has changed. Jan 10, 2017
@AmplabJenkins
Copy link

Can one of the admins verify this patch?

@hustfxj
Copy link
Contributor Author

hustfxj commented Jan 17, 2017

@srowen can you help review it ? I think it is a bug. Thank you very much.

@hustfxj
Copy link
Contributor Author

hustfxj commented Mar 9, 2017

@srowen @andrewor14 can you review it again? Thank you

@jiangxb1987
Copy link
Contributor

@hustfxj Unluckily we don't support multi-master nodes in standalone mode, so could you please close this PR? Thank you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants