[ https://issues.apache.org/jira/browse/CASSANDRA-8161?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14185373#comment-14185373 ]
Jonathan Ellis commented on CASSANDRA-8161: ------------------------------------------- We're talking about SSTable.acquireReference, right? That should require an awful lot of concurrency to livelock. How are people starting up that many concurrent repair sessions? > Spin loop when streaming tries to acquire SSTable references > ------------------------------------------------------------ > > Key: CASSANDRA-8161 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8161 > Project: Cassandra > Issue Type: Bug > Reporter: Yuki Morishita > Priority: Minor > > When streaming files, stream session tries to acquire references on SSTables > to stream. When there are multiple sessions trying to stream same part of > SSTables, they sometimes wait a *very* long time (nealy dead lock) on trying > to acquire references. > I think we need something we've done to AtomicSortedColulmns in > CASSANDRA-7546 to reduce contention. -- This message was sent by Atlassian JIRA (v6.3.4#6332)