Aggarwal-Raghav commented on PR #472: URL: https://github.com/apache/tez/pull/472#issuecomment-4147664286
CC @abstractdog , This jira is more for discussion/organization of tez docker image. **Key changes:** 1. Unified Docker Image: A single Dockerfile that builds an image for both AM and Task process. 2. Added a main entrypoint.sh that decide execution to either am-entrypoint.sh or child-entrypoint.sh based on the TEZ_COMPONENT environment variable. 3. [TEZ-4682](https://issues.apache.org/jira/browse/TEZ-4682) docker image i.e. `apache/tez-am:1.0.0-SNAPSHOT` -> `apache/tez:1.0.0-SNAPSHOT`. IMO we should publish the new one as part of [TEZ-4698](https://issues.apache.org/jira/browse/TEZ-4698), will help the consumer to not switch to new one. **Points to discuss:** (Its per my understanding and can be wrong) 1. As TEZ Child needs 5 java args: `amHost, amPort, containerId, token, attemptNo`. tez-am and tez-child are standalone docker process and tez-child will have to watch/read from zk `/tez-external-sessions/tez_am/server` to get this info. So, we would need to extend that capability like TEZ-4007 in TezChild as well, this is my understanding <img width="823" height="413" alt="image" src="https://github.com/user-attachments/assets/c65e481e-8370-4688-a1cd-7cca56208679"; /> 2. The `TaskScheduler` and `ContainerLauncher` needs to be extended i.e. new implementation such that TezAM can start TezChild, [TEZ-4665](https://issues.apache.org/jira/browse/TEZ-4665) needs to be completed before this. Once TEZ-4665 is done, we can get rid of `tez.local.mode` from docker `tez-site.xml` 3. Once TEZ-4665 + TEZ-4700 is done, [TEZ-4663](https://issues.apache.org/jira/browse/TEZ-4663) will be automatically done. The [ExternalAmWordCount.java ](https://github.com/apache/tez/blob/master/tez-examples/src/main/java/org/apache/tez/examples/ExternalAmWordCount.java) should work. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
