Skip to content

fix the config yarn.queue not take effect bug#3

Closed
velee wants to merge 1 commit intoapache:0.9.0from
velee:0.9.0
Closed

fix the config yarn.queue not take effect bug#3
velee wants to merge 1 commit intoapache:0.9.0from
velee:0.9.0

Conversation

@velee
Copy link

@velee velee commented Jun 18, 2015

No description provided.

@yanfang724
Copy link
Contributor

can you open a JIRA for this if this is still an issue for you? Because usually we do not take care of the pull requests. Thank you. Basic step is listed here http://samza.apache.org/contribute/rules.html . Thank you!

@navina
Copy link
Contributor

navina commented Sep 28, 2016

@velee Your patch has been committed in SAMZA-767.
Can you close this PR?

Thanks!

@velee velee closed this Oct 8, 2016
@velee velee deleted the 0.9.0 branch October 8, 2016 06:01
pdu-mn1 added a commit to pdu-mn1/samza that referenced this pull request Aug 6, 2018
Add static maps from tableId -> table/callback executors to share
executors among tasks in the same container.
khandelwal-ayush pushed a commit to khandelwal-ayush/samza that referenced this pull request Jun 12, 2024
…ployment while job is already running from previous deployment. (apache#3)

Live deployment - reduce job downtime by spawning a new AM for new deployment while job is already running from previous deployment.

This changeset introduces logic for the live deployment feature in which we spawn a new AM while the existing deployment of the job is still running. However we want to make sure that at any given time there are only one set of containers running for the job. The new AM comes up and when it is ready to spawn up the containers, it kills existing deployment of the job (which would kill existing AM and existing containers for the job), waits for it to be moved to KILLED state and then continues with requesting containers from RM.

This feature is behind a flag which is based on config of the job.

Other changes in this changeset include -
1: change in clientHelper code so that it returns active app ids for a job in sorted order based on start time. Earlier this method would have returned only one job always, since there is only on job in running state. However now it can return two when the job is being deployed.
2: Small refactor in SamzaYarnAppMasterService code to move “write of AM info on coordinator stream” to a new method and use that method if the live deployment feature flag is not set, which means there is no change in flow if "live deployment" flag is not set. If live deployment flag is set, then we write AM info on coordinator stream after the existing AM and all containers are killed.

---------

Co-authored-by: Rahul Yadav <rahyadav@rahyadav-ld2.linkedin.biz>
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