virajjasani commented on pull request #3711: URL: https://github.com/apache/hadoop/pull/3711#issuecomment-976586442
Starttime calculation for Namenode and Router seem better than Datanode case, because it's not the main classes that have starttime defined (i.e. NameNode and DFSRouter) but rather other classes that are internally initialized (e.g. FSNamesystem and Router) as part of the main process initialization, have the starttime defined. In the case of Datanode, the main class itself (i.e. DataNode) has starttime defined but it is initialized before soon as the Datanode is instantiated. Hence, I think we can improve only Datanode's starttime to match with the actual process start time. I understand that for a healthy process initialization, there should not be much difference in starttime if we keep it at instance init level (without this patch) vs if we keep it soon after RPC server is started (with this patch), but my concern is that if for some reason, RPC handler start time is significantly delayed, then we would have incorrect starttime reported. -- 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]
