[
https://issues.apache.org/jira/browse/CASSANDRA-10848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15403925#comment-15403925
]
Sylvain Lebresne commented on CASSANDRA-10848:
----------------------------------------------
As the test we're talking about here are paging tests, one thing that could be
worth a double check is CASSANDRA-10880. That is, we should make sure that
during upgrades from 2.X to 3.0.x, we make sure the native protocol version is
fixed to version 3 (we should not be using version 4 *even* when talking to 3.0
nodes, and this until the upgrade is over). If it's not guaranteed by the
tests, bad things will happen, including server side error that would break
query and get the {{Requested pages were not delivered before timeout}} error
above.
I'm mentioning that because I ran into this with another test today, so it
seems that by default the python driver will try to connect with the max
possible version to each node. I'm not familiar enough with the upgrade test
code enough to say if it fixes the protocol version or not, but if it doesn't,
it should.
> Upgrade paging dtests involving deletion flap on CassCI
> -------------------------------------------------------
>
> Key: CASSANDRA-10848
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10848
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jim Witschey
> Assignee: DS Test Eng
> Labels: dtest
> Fix For: 3.0.x, 3.x
>
>
> A number of dtests in the {{upgrade_tests.paging_tests}} that involve
> deletion flap with the following error:
> {code}
> Requested pages were not delivered before timeout.
> {code}
> This may just be an effect of CASSANDRA-10730, but it's worth having a look
> at separately. Here are some examples of tests flapping in this way:
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
