[
https://issues.apache.org/jira/browse/CASSANDRA-14051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16270109#comment-16270109
]
Kurt Greaves edited comment on CASSANDRA-14051 at 11/29/17 5:25 AM:
--------------------------------------------------------------------
AFAICT {{add_dc_after_mv_network_replication_test}} should have been failing
since CASSANDRA-12681 because it relies on adding a DC to the RF before adding
a node in that DC, and then it relies on that node bootstrapping and streaming
from the existing DC and having the correct data. TBH CASSANDRA-12681 was not a
good solution, it's quite useful to be able to set RF for a non-existent DC. I
think it's still necessary to keep the previous behaviour available, in which
case we could just add a "force\ignore" flag to alter/create if you want to add
a DC that doesn't exist i.e maybe something like {{CREATE KEYSPACE ... WITH
replication = {...} UNSAFE}}.
Or we could just revert CASSANDRA-12681 because it's not really helpful anyway.
It's not like we verify the replication factor or stop people from using
SimpleStrategy. Changing RF is a dangerous operation but still needs to be
flexible. I don't think this is actually a significant pain point, any mistakes
are quickly rectifiable, and I haven't really ever heard of anyone saying this
is a problem.
was (Author: kurtg):
AFAICT {{add_dc_after_mv_network_replication_test}} should have been failing
since CASSANDRA-12681 because it relies on adding a DC to the RF before adding
a node in that DC, and then it relies on that node bootstrapping and streaming
from the existing DC and having the correct data. TBH CASSANDRA-12681 was not a
good solution, it's quite useful to be able to set RF for a non-existent DC. I
think it's still necessary to keep the previous behaviour available, in which
case we could just add a "force\ignore" flag to alter if you want to add a DC
that doesn't exist.
> Many materialized_views_test are busted
> ---------------------------------------
>
> Key: CASSANDRA-14051
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14051
> Project: Cassandra
> Issue Type: Bug
> Components: Testing
> Reporter: Michael Kjellman
> Assignee: Kurt Greaves
>
> Many materialized_views_test are busted... For now we should disable the
> entire MV test suite until effort is put into making these test usable and
> helpful. Currently they aren't helpful and almost all fail.
> test_base_column_in_view_pk_commutative_tombstone_without_flush -
> materialized_views_test.TestMaterializedViews
> test_base_column_in_view_pk_complex_timestamp_with_flush -
> materialized_views_test.TestMaterializedViews
> add_dc_after_mv_network_replication_test -
> materialized_views_test.TestMaterializedViews
> add_dc_after_mv_simple_replication_test -
> materialized_views_test.TestMaterializedViews
> add_node_after_mv_test - materialized_views_test.TestMaterializedViews
> add_node_after_very_wide_mv_test -
> materialized_views_test.TestMaterializedViews
> add_node_after_wide_mv_with_range_deletions_test -
> materialized_views_test.TestMaterializedViews
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
