[ https://issues.apache.org/jira/browse/CASSANDRA-9832?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14711011#comment-14711011 ]
Sam Tunnicliffe commented on CASSANDRA-9832: -------------------------------------------- It was initially deemed necessary as an early version of the new API would hold a single OpOrder while processing an entire partition during compaction (& cleanup). In the version which was ultimately committed however, each row within a partition is updated as a distinct operation. So this would likely still be of benefit to the indexing subsystem if/when we come to improve the compaction & cleanup processing to batch row updates a bit more intelligently. As it stands though this is no longer critical, at least not for secondary indexing. > OpOrder: BarrierJumpingOp > ------------------------- > > Key: CASSANDRA-9832 > URL: https://issues.apache.org/jira/browse/CASSANDRA-9832 > Project: Cassandra > Issue Type: Improvement > Components: Core > Reporter: Benedict > Assignee: Benedict > Priority: Minor > Fix For: 3.0 beta 2 > > > Necessary for safe secondary index operation under the new API: a long > running operation for use with OpOrder, that can jump between barriers at > safe moments, without incurring the cost of starting/stopping except when > necessary. -- This message was sent by Atlassian JIRA (v6.3.4#6332)