[
https://issues.apache.org/jira/browse/CASSANDRA-10142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14705630#comment-14705630
]
Blake Eggleston edited comment on CASSANDRA-10142 at 8/20/15 7:49 PM:
----------------------------------------------------------------------
The problem appeared in commit {{db782362aaf25214a94c393b20cdbade829c5291}, as
part of CASSANDRA-6717. Although from what I've been able to tell so far, it's
just exposing a problem with the C* protocol frozen type protocol serialization
implementation, which is encoding the frozen map length as a 4 bytes
int,instead of the 2 expected by versions 1&2
was (Author: bdeggleston):
The problem appeared in commit {{db782362aaf25214a94c393b20cdbade829c5291}, as
part of CASSANDRA-6717. Although from what I've been able to tell so far, it's
just exposing a problem with the python driver, which isn't correctly
deserializing the (now) frozen maps from the schema tables in protocol versions
1 & 2.
> Driver cannot connect with protocol version < 3
> -----------------------------------------------
>
> Key: CASSANDRA-10142
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10142
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jim Witschey
> Assignee: Blake Eggleston
> Fix For: 3.0 beta 1
>
> Attachments: repro.sh
>
>
> When trying to connect with the Python driver to a cluster running trunk, the
> connection fails with a {{NoHostAvailable}} exception if the protocol version
> is set to 1 or 2.
> The attached script reproduces the error and demonstrates that the connection
> works if the protocol version is 3 or 4.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
