[ https://issues.apache.org/jira/browse/HIVE-17841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16210612#comment-16210612 ]
Sergey Shelukhin edited comment on HIVE-17841 at 10/19/17 5:40 AM: ------------------------------------------------------------------- This preliminary patch is based on master although it will be committed after the pools patch. I'm pretty sure tests will fail... will look tomorrow. This rewrites WM to use a single thread to process almost everything that can happen to a session that is in use (so, excluding out of bounds expiration). This is because there are too many things coming from multiple sources and otherwise synchronization and reasoning about all the things happening in parallel becomes a huge pain. Also makes some things like getting session from the pool, waiting for AM registry, etc. async to facilitate that. There are some followup jiras that would be needed - handling internal AM restart (AM registry info update), and LLAP cluster information changes (the logic in WM is simple, the information propagation needs to be added). cc [~prasanth_j] was (Author: sershe): This preliminary patch is based on master although it will be committed after the pools patch. I'm pretty sure tests will fail... will look tomorrow. This rewrites WM to use a single thread to process almost everything that can happen to a session that is in use (so, excluding out of bounds expiration). There are some followup jiras that would be needed - handling internal AM restart (AM registry info update), and LLAP cluster information changes (the logic in WM is simple, the information propagation needs to be added). > implement applying the resource plan > ------------------------------------ > > Key: HIVE-17841 > URL: https://issues.apache.org/jira/browse/HIVE-17841 > Project: Hive > Issue Type: Sub-task > Reporter: Sergey Shelukhin > Assignee: Sergey Shelukhin > Attachments: HIVE-17841.patch > > -- This message was sent by Atlassian JIRA (v6.4.14#64029)