singer-bin commented on PR #8573: URL: https://github.com/apache/hadoop/pull/8573#issuecomment-4861832616
Update: it turns out I do not have permission to edit the Release Note field on YARN-11967. Could a committer please set the following release note when merging? Thanks! ``` Previously, setting yarn.nodemanager.resource.memory.enforced (default: true) alone caused the NodeManager to skip the polling-based physical/virtual memory check, assuming the CGroups OOM killer would enforce the limit. However, the CGroups memory hard limit is only applied when yarn.nodemanager.resource.memory.enabled (default: false) is also enabled. As a result, with default settings the polling-based memory check was skipped even though no CGroups memory limit was actually in effect, allowing containers that exceeded their memory request to run unbounded. With this change, strict memory enforcement (which skips the polling check) takes effect only when BOTH yarn.nodemanager.resource.memory.enabled and yarn.nodemanager.resource.memory.enforced are true. Clusters that do not enable CGroups memory but rely on yarn.nodemanager.pmem-check-enabled / yarn.nodemanager.vmem-check-enabled will now correctly kill containers that exceed their limits via the polling mechanism. Clusters that already enable CGroups memory are unaffected. No configuration change is required. ``` -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
