[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jay Kreps resolved KAFKA-1008.
--
Resolution: Fixed
Checked in on 0.8 and trunk.
> Unmap before resiz
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13791171#comment-13791171
]
Jun Rao commented on KAFKA-1008:
Thanks for the patch for trunk. +1.
> Unmap
to trunk and
also adds a fix for the windows compatibility issue in KAFKA-1036 in
OffsetCheckpoint.scala.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
>
n v8.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, log
>Affects Vers
't think it is very safe to have methods whose correctness
depends on the existing call pattern. The synchronization doesn't hurt, in any
case since this is not in any critical read or write path.
> Unmap before resizing
> -
>
80. OffsetIndex: The patch synchronizes in readLastOffset(), is that necessary?
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
>
ll.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, log
>Aff
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13773517#comment-13773517
]
Sriram Subramanian commented on KAFKA-1008:
---
+1
> Unmap
I intend
inLock(x) {
foo
}
to be read as "in lock x do foo". So I like it since it is declarative.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/ji
bout cygwin.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, lo
hink).
3. I intend
inLock(x) {
foo
}
to be read as "in lock x do foo". So I like it since it is declarative.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jay Kreps updated KAFKA-1008:
-
Attachment: KAFKA-0.8-1008-v8.patch
> Unmap before resiz
you address Jun's review comments and see if
we can resolve this JIRA? This is marked for the 0.8 final release
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-10
sive.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, lo
g on Windows. Please check it in.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Compon
.scala
method - write and read
3. Is autolock a better name than inlock?
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project:
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jay Kreps updated KAFKA-1008:
-
Attachment: KAFKA-0.8-1008-v7.patch
Attached a version rebased to 0.8.
> Unmap bef
at
java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1110)
at
java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:603)
at java.lang.Thread.run(Thread.java:722)
> Unmap before resizing
> -
>
>
fix some of the issues in Jay's patch.
Please check if it works!
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: KAFKA-trunk-1008-v7.patch
> Unmap before resiz
; Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, log
>Affects Versions: 0.8
>
gher order function makes the code look a
lot cleaner!
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
>
tested on Windows. I'm not
sure if there is anyone who has access to a Windows machine and could reproduce
the old problem who could verify that this patch fixes it?
> Unmap before resizing
> -
>
> Key: KAFKA-1008
&
's a major blocker for anyone that
wants to use Kafka on windows
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: unmap-v5.patch
> Unmap before resiz
see if it reflects your suggestion. I'm still
learning Scala, so please provide any feedback!
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
>
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: (was: unmap-v4.patch)
> Unmap before resiz
rced mmap clenaing all be inside the isWindows check. Thisshould fix things
on windows and not functionally change perf at all on linux (which is good).
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https:
eads. Do you want to file a separate JIRA
to address the lock read issue?
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
>
ding to this random page):
http://www.javaneverdie.com/java/java-os-name-property-values/
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
>
ection capabilities or
other OS-specific functionality.
Then elsewhere we can just do
if(Os.isWindows) lock.lock()
and the corresponding unlock.
Does that seem like it would work?
> Unmap before resizing
> -
>
> Key
a shallow copy of the buffer. Do you have a case
where the buffer is being used while resizing? I did a quick grep, and it looks
like resize is only being used during logfile loading or truncation.
> Unmap before resizing
> -
>
>
cate()
(orig.asInstanceOf[sun.nio.ch.DirectBuffer]).cleaner().clean()
copy.get()
My suspicion is that something terrible will happen, but I could be wrong.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
>
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: unmap-v4.patch
> Unmap before resiz
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: (was: unmap-v3.patch)
> Unmap before resiz
the exceptions.
I found the following coding pattern
val idx = mmap.duplicate
in this file. It seems like it always makes a copy of the buffer for "read".
> Unmap before resizing
> -
>
> Key: KAFKA-1008
&g
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: (was: unmap-v2.patch)
> Unmap before resiz
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: unmap-v3.patch
> Unmap before resiz
Launcher.java:301)
at java.lang.ClassLoader.loadClass(ClassLoader.java:247)
... 1 more
We also can't really be too sure what kind of exceptions clean() might throw.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL:
chronized,
which should prevent multiple threads trying to resize/unmap the files.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
>
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: unmap-v2.patch
> Unmap before resiz
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: (was: unmap-v1.patch)
> Unmap before resiz
07 AM, lizziew wrote:
>
> > GitHub user lizziew opened a pull request:
> >
> > https://github.com/apache/kafka/pull/6
> >
> > Unmap before resizing
> >
> > While I was studying how MappedByteBuffer works, I sa
x27;t think so...).
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components
s no-op for now; we can add support for
other JVM environments later on.
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
>
only environments?
> Unmap before resizing
> -
>
> Key: KAFKA-1008
> URL: https://issues.apache.org/jira/browse/KAFKA-1008
> Project: Kafka
> Issue Type: Bug
> Components: core, lo
t 12:07 AM, lizziew wrote:
> GitHub user lizziew opened a pull request:
>
> https://github.com/apache/kafka/pull/6
>
> Unmap before resizing
>
> While I was studying how MappedByteBuffer works, I saw a sharing
> runtime exception on Windows. I applied what I learned t
[
https://issues.apache.org/jira/browse/KAFKA-1008?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Elizabeth Wei updated KAFKA-1008:
-
Attachment: unmap-v1.patch
Patch file
> Unmap before resiz
Elizabeth Wei created KAFKA-1008:
Summary: Unmap before resizing
Key: KAFKA-1008
URL: https://issues.apache.org/jira/browse/KAFKA-1008
Project: Kafka
Issue Type: Bug
Components
apache/kafka/pull/6
>
> Unmap before resizing
>
> While I was studying how MappedByteBuffer works, I saw a sharing
> runtime exception on Windows. I applied what I learned to generate a patch
> which uses an internal open JDK API to solve this problem.
>
> ---
GitHub user lizziew opened a pull request:
https://github.com/apache/kafka/pull/6
Unmap before resizing
While I was studying how MappedByteBuffer works, I saw a sharing runtime
exception on Windows. I applied what I learned to generate a patch which uses
an internal open JDK API
51 matches
Mail list logo
