[ https://issues.apache.org/jira/browse/HIVE-14172?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15371527#comment-15371527 ]
Sergey Shelukhin commented on HIVE-14172: ----------------------------------------- Hmm, I thought I commented here. Probably commented on some wrong JIRA 0_o [~gopalv] ping? Failures are either known or caused by NN being in safe mode. > LLAP: force evict blocks by size to handle memory fragmentation > --------------------------------------------------------------- > > Key: HIVE-14172 > URL: https://issues.apache.org/jira/browse/HIVE-14172 > Project: Hive > Issue Type: Bug > Reporter: Nita Dembla > Assignee: Sergey Shelukhin > Attachments: HIVE-14172.01.patch, HIVE-14172.patch > > > In the long run, we should replace buddy allocator with a better scheme. For > now do a workaround for fragmentation that cannot be easily resolved. It's > still not perfect but works for practical ORC cases, where we have the > default size and smaller blocks, rather than large allocations having trouble. -- This message was sent by Atlassian JIRA (v6.3.4#6332)