KevinWikant commented on PR #4568: URL: https://github.com/apache/hadoop/pull/4568#issuecomment-1205781950
Hey Xiaohui, appreciate you taking the time to review this :) Just wanted to clarify a few things: ``` This is a complete unnecessary change. It was mentioned in your last JIRA [HDFS-16064](https://issues.apache.org/jira/browse/HDFS-16064) by [yanbin.zhang](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=it_singer) that you are finding a wrong root cause in that JIRA and you didnt coordinate or reply them ? ``` I did reply to yanbin.zhang in the JIRA, you can see my comment just under theirs. I would also add: a) the root cause is correct as per the testing that was conducted b) I have seen this issue (i.e. HDFS-16064) impact production Hadoop clusters ``` This current change seems to unnecessary. If it would be a real production bug , someone in community would have experienced it. It is not like that you are reading the code someday and you identified some bug like this. Block management is very much critical and unnecessary changes always lead to more critical bugs... ``` I understand that there is no clear reporting of this issue beforehand by the community. That being said, the behavior is reproducible (as per the testing details in this PR/JIRA) and the behavior results in block invalidation requests (i.e. DNA_INVALIDATE) sent to datanodes failing due to incorrect GenerationStamp. I would argue that sending an incorrect GenerationStamp in a DNA_INVALIDATE request is a bug because it causes the request to fail when it otherwise could have succeeded if the correct GenerationStamp is used. I would ask that you more clearly articulate why this change is un-necessary. Do you believe that sending an incorrect GenerationStamp in a DNA_INVALIDATE request causing the request to fail is not a bug? Or do you believe that this reproducible behavior should not be addressed simply because it has not been frequently reported as production impacting by the community? -- 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]
