[
https://issues.apache.org/jira/browse/CASSANDRA-7581?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15278153#comment-15278153
]
Alex Petrov edited comment on CASSANDRA-7581 at 5/11/16 7:29 AM:
-----------------------------------------------------------------
This issue looks related to
[7592|https://issues.apache.org/jira/browse/CASSANDRA-7592], although 7592
seems to have a more complete scope description. Although suggested solution
might solve both issues: by dropping read requests coordinator will be able to
pick another node to read from.
-This problem seems not to apply to local mutations, as they're always checked
for {{appliesLocally}}. So in the end it might be a good idea to have similar
flow for reads.-
I've checked the code in more detail. At the moment both mutations and read
commands are applied without additional checks for token endpoints. It can be
done quite simply via the {{StorageService}}.
But in general, replica may have an incorrect view of the ring, just as
coordinator can. Although if the replica has the wrong view, it would be better
if it doesn't serve requests.
was (Author: ifesdjeen):
This issue looks related to
[7592|https://issues.apache.org/jira/browse/CASSANDRA-7592], although 7592
seems to have a more complete scope description. Although suggested solution
might solve both issues: by dropping read requests coordinator will be able to
pick another node to read from.
This problem seems not to apply to local mutations, as they're always checked
for {{appliesLocally}}. So in the end it might be a good idea to have similar
flow for reads.
> Stop read requests for partitions which don't belong to a node
> --------------------------------------------------------------
>
> Key: CASSANDRA-7581
> URL: https://issues.apache.org/jira/browse/CASSANDRA-7581
> Project: Cassandra
> Issue Type: Bug
> Reporter: sankalp kohli
> Assignee: Alex Petrov
> Priority: Minor
>
> A C* instance will serve CF level reads with null values or old values to a
> co-ordinator even if the instance is not responsible for the data.
> This is bad if co-ordinator has a wrong view of the ring and will return
> wrong results back to the client.
> A solution to this problem could be to not respond or return error to such
> read requests.
> Tricky part here is to respond to read requests when the instance has just
> lost ownership for a few ring delay.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
