Make YarnAppRunner more robust to failure.
- Dominant language
- Java
- Stars
- 200
- Forks
- 73
- PR merge metrics
- No merged PRs in 30d
Description
While reviewing #563 I noticed a more general issue. If any failure occurs before [YarnAppRunner.java line 232](https://github.com/fluo-io/fluo/blob/50d5fde1f9f47db07c313d9e7e6f6fa6fcf5a117/modules/cluster/src/main/java/io/fluo/cluster/runner/YarnAppRunner.java#L232) it could leave things in an inconsistent state where the Fluo command line tools do not work well.
It would be nice to make this more robust. One possible way to do this is to place a marker in zookeeper, before trying to start anything in Twill, that indicates that yarn startup is in progress. Other Fluo commands could use this to know if startup is in progress. Other Fluo command could offer users advice about running yarn commands if it seems that startup is taking to long.
Contributor guide
Assessment
This issue has not been assessed yet.