[
https://issues.apache.org/jira/browse/CASSANDRA-11521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15296244#comment-15296244
]
Benedict commented on CASSANDRA-11521:
--------------------------------------
Language is imprecise - I didn't mean a blanket ban on OpOrder, just that it
should not be used for the long-life cursor's sstable access, only held during
any active service of the query by the server. If we were to offer isolation,
the memtable contents applicable to the query would have to be copied anyway
(it could be referenced directly until flush, but given the current on/off heap
hybrid situation that probably isn't desirable) , which actually might be a
point against offering isolation.
I don't have time to remind myself of the intricacies of the issue with
referencing sstables, but I did see in passing a ticket that [~snazy]
filed/fixed for suppressing a very spamming log message with respect to failing
to reference sstables, and he may be able to point you to it. That's what I'm
referring to - the log messages are the symptom (and were deliberately not made
spam proof to make it obvious it was happening, since it's a bad thing). As
far as I recall, there's a period during the replacement of sstables after
compaction where the expired versions are visible and cannot possibly be
referenced and so the selection loop spins. It's probably quite easily fixed,
but you will have to investigate for yourself.
> Implement streaming for bulk read requests
> ------------------------------------------
>
> Key: CASSANDRA-11521
> URL: https://issues.apache.org/jira/browse/CASSANDRA-11521
> Project: Cassandra
> Issue Type: Sub-task
> Components: Local Write-Read Paths
> Reporter: Stefania
> Assignee: Stefania
> Fix For: 3.x
>
>
> Allow clients to stream data from a C* host, bypassing the coordination layer
> and eliminating the need to query individual pages one by one.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
