[ https://issues.apache.org/jira/browse/HBASE-17528?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16398114#comment-16398114 ]
stack commented on HBASE-17528: ------------------------------- Unscheduled from 2.0.0. Was thinking we deprecate Scan being able to set this in 2.0.0 and that later we'd just have the server figure it out for user. Meantime pread all the time. Years ago measuring the difference was little. Perhaps when stars are aligned and the DN is feeling frisky, it would start dumping us data by the bucket-full. But needs investigation. Also in need of investigation is WTF my logs are filled with slow sync LOGs all day long.... > Remove the Scan.ReadType flag introduced in HBASE-17045 > ------------------------------------------------------- > > Key: HBASE-17528 > URL: https://issues.apache.org/jira/browse/HBASE-17528 > Project: HBase > Issue Type: Task > Components: scan > Affects Versions: 2.0.0 > Reporter: Duo Zhang > Assignee: stack > Priority: Major > Attachments: HBASE-17528.branch-2.001.patch > > > It is used to keep the old behavior of small scan. We should have a mechanism > which selects pread or stream automatically at server side. And in general, > we have discussed many times whether it is OK to always use pread for a user > scan. -- This message was sent by Atlassian JIRA (v7.6.3#76005)