[ 
https://issues.apache.org/jira/browse/CASSANDRA-3868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13214535#comment-13214535
 ] 

Peter Schuller commented on CASSANDRA-3868:
-------------------------------------------

I completely agree about not adding more counter-specific code.

I still think it is a very useful feature to retain, for the same reasons as 
before. I'm not suggesting running with it in general, but it is an extremely 
effective pressure release valve for large data size counter clusters - keeping 
in mind in particular that counters are, in addition to being potentially very 
expensive due to reading, non-idempotent and unsafe to re-try. A cluster even 
"slightly" overloaded or misbehaving to the point of causing timeouts is 
effectively causing you to drop increments (or overcount if you "incorrectly" 
retry). In other words, it's fairly easy to end up in a position where you 
might prefer running with replicate on write turned off for a few minutes/hours 
while you fix something; much more so than with normal writes.

So in summary, the sensitivity to timeouts with counters from the client's 
perspective, in combination with the easy with which you can get into temporary 
capacity issues due to read-on-write, makes it, in my opinion, a useful thing 
to retain.

                
> Remove or nullify replicate_on_write option
> -------------------------------------------
>
>                 Key: CASSANDRA-3868
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-3868
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>    Affects Versions: 0.8.0
>            Reporter: Brandon Williams
>             Fix For: 1.1.0
>
>         Attachments: 3868.txt
>
>
> My understanding from Sylvain is that setting this option to false is rather 
> dangerous/stupid, and you should basically never do it.  So 1.1 is a good 
> time to get rid of it, or make it a no-op.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to