[
https://issues.apache.org/jira/browse/CASSANDRA-9303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15074033#comment-15074033
]
Stefania edited comment on CASSANDRA-9303 at 12/30/15 10:33 AM:
----------------------------------------------------------------
The hanging tests were caused by the way in which we run cqlsh in ccm, this
[pull request|https://github.com/pcmanus/ccm/pull/432] fixed it.
The remaining failures were caused by the following two things:
* Handling of temporary files is quite different on Windows, the changes to
address this are only in the dtest code, see the second commit of the [pull
request|https://github.com/riptano/cassandra-dtest/pull/724].
* The path names should have been normalized, see [this
commit|https://github.com/stef1927/cassandra/commit/295219dfbcf24ece9729030cce6e9638899b2842].
I've also changed a few more things, mostly discovered whilst trying to
reproduce CASSANDRA-10938 on Windows:
* Reverted to batching by replica to avoid Cassandra processes using too much
CPU. Batching by replica was changed to batching by partition key during the
code review of CASSANDRA-9302 because there is a cost in determining the
replicas of each record. However, sending batches with records on different
replicas is probably worst then spending a few cycles in Python determining the
correct replicas. It also allows up to use LOGGED batching, see next point.
* EDIT: -Changed batch type from UNLOGGED to LOGGED to avoid a WARN in the
Cassandra log files and for more consistent failed batch status reporting (even
though INSERT should be idempotent, so this can be changed back to UNLOGGED if
performance is impacted too much but it shouldn't since all parititions should
be local).- This was too bad for performance, see discussion on CASSANDRA-10938.
* Fixed a problem with cassandra-stress that only manifested on Windows and on
trunk when using a custom profile. However the Windows stress launch scripts
were incorrect from 2.1 onwards.
* EDIT: Added checks on invalid {{boolstyle}} options, as requested by dtest
code review.
I worked on the 2.2 patch and merged upwards. I also cherry-picked back to 2.1
with manual conflict resolution in bin/cqlsh. Even though we don't support
Windows for 2.1 I figured it was best to fix these problems anyway.
was (Author: stefania):
The hanging tests were caused by the way in which we run cqlsh in ccm, this
[pull request|https://github.com/pcmanus/ccm/pull/432] fixed it.
The remaining failures were caused by the following two things:
* Handling of temporary files is quite different on Windows, the changes to
address this are only in the dtest code, see the second commit of the [pull
request|https://github.com/riptano/cassandra-dtest/pull/724].
* The path names should have been normalized, see [this
commit|https://github.com/stef1927/cassandra/commit/295219dfbcf24ece9729030cce6e9638899b2842].
I've also changed a few more things, mostly discovered whilst trying to
reproduce CASSANDRA-10938 on Windows:
* Reverted to batching by replica to avoid Cassandra processes using too much
CPU. Batching by replica was changed to batching by partition key during the
code review of CASSANDRA-9302 because there is a cost in determining the
replicas of each record. However, sending batches with records on different
replicas is probably worst then spending a few cycles in Python determining the
correct replicas. It also allows up to use LOGGED batching, see next point.
* Changed batch type from UNLOGGED to LOGGED to avoid a WARN in the Cassandra
log files and for more consistent failed batch status reporting (even though
INSERT should be idempotent, so this can be changed back to UNLOGGED if
performance is impacted too much but it shouldn't since all parititions should
be local).
* Fixed a problem with cassandra-stress that only manifested on Windows and on
trunk when using a custom profile. However the Windows stress launch scripts
were incorrect from 2.1 onwards.
I worked on the 2.2 patch and merged upwards. I also cherry-picked back to 2.1
with manual conflict resolution in bin/cqlsh. Even though we don't support
Windows for 2.1 I figured it was best to fix these problems anyway.
> Match cassandra-loader options in COPY FROM
> -------------------------------------------
>
> Key: CASSANDRA-9303
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9303
> Project: Cassandra
> Issue Type: New Feature
> Components: Tools
> Reporter: Jonathan Ellis
> Assignee: Stefania
> Priority: Critical
> Fix For: 2.1.x, 2.2.x, 3.0.x, 3.x
>
> Attachments: dtest.out
>
>
> https://github.com/brianmhess/cassandra-loader added a bunch of options to
> handle real world requirements, we should match those.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
