If it's taking that long, you may be working on hardware (or a VM?) which
is too underpowered to run some of the tests reliably.
The README.md has instructions for how to build different components. In
your case you want
./gradlew core:test
Depending on your hardware, you may want to adjust the
On Wed, Feb 3, 2016 at 3:57 PM, Shane MacPhillamy
wrote:
> Hi
>
> I’m just coming up to speed with Kafka. Some beginner questions, may be
> point me to where I can find the answers please:
>
> 1. In a Kafka cluster what determines the maximum number of concurrent
>
Gary,
Here are a few concrete examples from Kafka and Confluent Platform:
JSON (baked into Kafka Connect, not specifically designed for standalone
serialization but they should work for that):
The default max message size is 1MB. You'll probably need to increase a few
settings -- the topic max message size on a per-topic basis on the broker
(or broker-wide with message.max.bytes), the max.partition.fetch.bytes on
the new consumer, etc. You need to make sure all of the producer, broker,
Daniel,
Awesome, Ruby folks could use more Kafka love! I added the library to the
clients list here:
https://cwiki.apache.org/confluence/display/KAFKA/Clients#Clients-Ruby I'm
also cc'ing this to the clients list since I think they'd be interested as
well.
Lots of folks are using the Java
Sunny,
As I said on Twitter, I'm stoked to hear you're working on a Mongo
connector! It struck me as a pretty natural source to tackle since it does
such a nice job of cleanly exposing the op log.
Regarding the problem of only getting deltas, unfortunately there is not a
trivial solution here --
ou get the current value not
> every necessarily every intermediate) but that should be okay for most
> uses.
>
> -Jay
>
> On Fri, Jan 29, 2016 at 8:54 AM, Ewen Cheslack-Postava <e...@confluent.io>
> wrote:
>
> > Sunny,
> >
> > As I said on Twitter, I'm stoked t
On Fri, Jan 29, 2016 at 7:06 AM, Randall Hauch <rha...@gmail.com> wrote:
> On January 28, 2016 at 7:07:02 PM, Ewen Cheslack-Postava (
> e...@confluent.io) wrote:
>
> Randall,
>
> Great question. Ideally you wouldn't need this type of state since it
> should really
Randall,
Great question. Ideally you wouldn't need this type of state since it
should really be available in the source system. In your case, it might
actually make sense to be able to grab that information from the DB itself,
although that will also have issues if, for example, there have been
If you don't shut it down properly and there are outstanding requests (e.g.
if you call producer.send() and don't call get() on the returned future),
then you could potentially lose data. Calling producer.close() flushes all
the data before returning, so shutting down properly ensures no data will
One option is to instantiate and invoke the DefaultPartitioner yourself (or
whatever partitioner you've specified for partitioner.class). However, that
will require passing in a Cluster object, which you'll need to construct
yourself. This is just used to get the number of partitions for the topic
It's not an iterator (ConsumerRecords is a collection of records), but you
also won't just get the entire set of messages all at once. You would have
the same issue if you set auto.offset.reset to earliest for a new consumer
-- everything that's in the topic will need to be consumed.
Under the
No, you don't need to keep adding ZK nodes. You should have a 3 or 5 node
ZK cluster. The more nodes you use, the slower write performance becomes,
so adding more can hurt performance of any ZK-related operations. The
tradeoff between 3 and 5 ZK nodes is fault tolerance (better with 5) vs
write
Nikhil,
You should search the mailing list archives, but I'm not aware of any
discussion around that. If you wanted to try something like that, you might
be able to accomplish it via FUSE or similar. For example, this page lists
ways you can mount HDFS as a normal filesystem, including fuse-based
Hi,
The new consumer is single threaded. You can layer multi-threaded
processing on top of it, but you'll definitely need to be careful about how
offset commits are handled to ensure a) processing of a message is actually
*complete* not just passed off to another thread before committing an
Is the consumer registration failing, or the subsequent calls to read from
the topic? From the error, it sounds like the latter -- a conflict during
registration should generate a 40902 error.
Can you give more info about the sequence of requests that causes the
error? The set of commands you
t; have yet to find the right one.
>
> Thanks!
>
> Andrew
>
> On Fri, Jan 8, 2016 at 10:54 AM, Ewen Cheslack-Postava <e...@confluent.io>
> wrote:
>
> > Andrew,
> >
> > kafka-producer-perf-test.sh is just a wrapper around
> > orga.apache.kafka.c
-class.sh
> org.apache.kafka.clients.tools.ProducerPerformance
> is single threaded? Or is there any way to specify number of threads?
>
> On Fri, Jan 8, 2016 at 1:24 PM, Ewen Cheslack-Postava <e...@confluent.io>
> wrote:
>
> > Ah, sorry, I missed the version number in your title. I think this too
Andrew,
kafka-producer-perf-test.sh is just a wrapper around
orga.apache.kafka.clients.tools.ProducerPerformance and all command line
options should be forwarded. Can you just pass a --producer-props to set
max.request.size to a larger value?
-Ewen
On Fri, Jan 8, 2016 at 7:51 AM, Andrej
ET client, but in the proxy it
> doesn’t appear to work like that.
>
> -Original Message-
> From: Ewen Cheslack-Postava [mailto:e...@confluent.io]
> Sent: Thursday, January 07, 2016 1:18 PM
> To: users@kafka.apache.org
> Subject: Re: Kafka-Rest question
>
> On Thu, Jan 7, 2016 at 12
Looks good! I've added it to the clients page here:
https://cwiki.apache.org/confluence/display/KAFKA/Clients
-Ewen
On Thu, Jan 7, 2016 at 9:12 AM, Dana Powers wrote:
> Very nice!
> On Jan 7, 2016 04:41, "Pierre-Yves Ritschard" wrote:
>
> > Hi list,
>
Chandra,
If its a separate app to collect logs, it would presumably run on the same
server IIS is running on since that's where logs would be generated.
-Ewen
On Tue, Dec 29, 2015 at 9:22 PM, chandra sekar
wrote:
> Dear Ewen,
> Where do i run the separate
Chandra,
If you're just serving files from IIS and want to collect logs, you'll
probably want to run a separate application to collect the log files and
report each log entry to Kafka.
If you're running a web application, you can use the producer yourself to
report events to Kafka.
-Ewen
On
is if you want to be able to mix consumers using different
libraries in the same consumer group (consumers in different groups using
different libraries should always be fine).
-Ewen
On Mon, Dec 28, 2015 at 4:16 PM, Ewen Cheslack-Postava <e...@confluent.io>
wrote:
> Yes, version=0 should
-----
>
> Maybe I misunderstood the purpose of this version field?
>
> On Thu, 24 Dec 2015 at 00:27 Ewen Cheslack-Postava <e...@confluent.io>
> wrote:
>
> > Oleksiy,
> >
> > Where are you specifying the version? Unless I'm missing something, th
Oleksiy,
Where are you specifying the version? Unless I'm missing something, the
JoinGroup protocol doesn't include versions so I'm not sure I understand
the examples you are giving. Are the version numbers included in the
per-protocol metadata?
You can see exactly how the consumer coordinator
Dana,
Not sure about the old merge script, but with the new one used for GitHub
PRs it tracks the branches you choose to commit to and can update the JIRA
automatically, tagging the appropriate fix versions. So from now on, the
appropriate JIRA queries for issues with, e.g., fix version 0.9.0.1
What version of Gradle are you using and can you give the exact command
you're running?
-Ewen
On Wed, Dec 23, 2015 at 5:49 PM, Oliver Pačut
wrote:
> Hello,
>
> I am having trouble using Gradle to build Kafka. I get the error:
>
>
> FAILURE: Build failed with an
producer with 0.8.1.1 brokers without problems.
> > Version of scala matters if you are building with scala or some other
> > components that use scala.
> > Hope this helps.
> >
> > --
> > Andrey Yegorov
> >
> > On Wed, Dec 23, 2015 at 1:11 PM, Ew
Mark,
There are definitely limitations to using JDBC for change data capture.
Using a database-specific implementation, especially if you can read
directly off the database's log, will be able to handle more situations
like this. Cases like the one you describe are difficult to address
uent.io/2.0.0/connect/devguide.html
>
> Thanks,
> Roman
>
>
>
> On Wednesday, November 11, 2015 6:59 AM, Ewen Cheslack-Postava <
> e...@confluent.io> wrote:
> Hi Venkatesh,
>
> If you're using the default settings included in the sample configs, it'll
> expect
Meghana,
It looks like this functionality was removed in
https://issues.apache.org/jira/browse/KAFKA-1650, although I don't see
explicit discussion of the issue in the JIRA so I'm not sure of the exact
motivation. Maybe Becket or Guozhang can offer some insight about if it is
necessary (that JIRA
Svante,
Just to clarify, the HDFS connector relies on some Avro translation code
which is in a separate repository. You need the
https://github.com/confluentinc/schema-registry repository built before the
kafka-connector-hdfs repository to get that dependency.
Confluent has now also released
Kashif,
The difference is that close() will also shut down the producer such that
it can no longer send any messages. flush(), in contrast, is useful if you
want to make sure that all the messages enqueued so far have been sent and
acked, but also want to send more messages after that.
-Ewen
On
jar:0.9.0.0,
> io.confluent:kafka-connect-avro-converter:jar:2.0.0-SNAPSHOT,
> io.confluent:common-config:jar:2.0.0-SNAPSHOT: Could not find artifact
> org.apache.kafka:connect-api:jar:0.9.0.0 in confluent
>
> On Thu, Nov 12, 2015 at 2:59 PM, Ewen Cheslack-Postava <e...@confluent.io&g
Hi Adaryl,
First, it looks like you might be trying to use the old producer interface.
That interface is going to be deprecated in favor of the new producer
(under org.apache.kafka.clients.producer). I'd highly recommend using the
new producer interface instead.
Second, perhaps this repository
he sink as part of a confluent
> project (
>
> https://github.com/confluentinc/copycat-hdfs/blob/master/src/main/java/io/confluent/copycat/hdfs/HdfsSinkConnector.java
> ).
> Does it mean that I build this project and add the jar to kafka libs ?
>
>
>
>
> On Tue, Nov
Passing 0 as the port should let you do this. This is how we get the tests
to work without assuming a specific port is available. The
KafkaServer.boundPort(SecurityProtocol) method can be used to get the port
that was bound.
-Ewen
On Tue, Nov 10, 2015 at 11:23 PM, Hemanth Yamijala
Jun, not sure if this is just because of the RC vs being published on the
site, but the links in the release notes aren't pointing to
issues.apache.org. They're relative URLs instead of absolute.
-Ewen
On Tue, Nov 10, 2015 at 3:38 AM, Flavio Junqueira wrote:
> -1 (non-binding)
Hi Venkatesh,
If you're using the default settings included in the sample configs, it'll
expect JSON data in a special format to support passing schemas along with
the data. This is turned on by default because it makes it possible to work
with a *lot* more connectors and data storage systems
in
> double-quotes.
> 2) Once I hit the above JsonParser error on the SinkTask, the connector is
> hung, doesn't take any more messages even proper ones.
>
>
> On Tue, Nov 10, 2015 at 1:59 PM, Ewen Cheslack-Postava <e...@confluent.io>
> wrote:
>
> > Hi Venka
0.9.0.0 is not released yet, but the last blockers are being addressed and
release candidates should follow soon. The docs there are just staged as we
prepare for the release (note, e.g., that the latest release on the
downloads page http://kafka.apache.org/downloads.html is still 0.8.2.2).
-Ewen
If you want the full round-trip latency, you need to measure that at the
client. The performance measurement tools should do this pretty accurately.
For example, if you just want to know how long it takes to produce a
message to Kafka and get an ack back, you can use the latency numbers
reported
You can accomplish this with the console consumer -- it has a formatter
flag that lets you plug in custom logic for formatting messages. The
default does not do any formatting, but if you write your own
implementation, you just need to set the flag to plug it in.
You can see an example of this in
There are a couple of advantages. First, scalability. Writes to Kafka are
cheaper than writes to ZK. Kafka-based offset storage is going to be able
to handle significantly more consumers (and scale out as needed since
writes are spread across all partitions in the offsets topic). Second, once
you
Yes, 0.9 will include the new consumer.
On Tue, Oct 13, 2015 at 12:50 PM, Rajiv Kurian wrote:
> A bit off topic but does this release contain the new single threaded
> consumer that supports the poll interface?
>
> Thanks!
>
> On Mon, Oct 12, 2015 at 1:31 PM, Jun Rao
Not sure if I'd call it a blocker, but if we can get it in I would *really*
like to see some solution to
https://issues.apache.org/jira/browse/KAFKA-2397 committed. Without an
explicit leave group, even normal operation of groups can leave some
partitions unprocessed for 30-60s at a time under
On Javadocs, both new clients (producer and consumer) have very thorough
documentation in the javadocs. 0.9.0.0 will be the first release with the
new consumer.
On deserialization, the new consumer lets you specify deserializers just
like you do for the new producer. But the old consumer supports
ConsumerConnector is part of the old consumer API (which is what is
currently released; new consumer is coming in 0.9.0). That class is not in
kafka-clients, it is in the core Kafka jar, which is named with the Scala
version you want to use, e.g. kafka_2.10.
-Ewen
On Thu, Oct 8, 2015 at 1:24 PM,
And you can get the current assignment in the new consumer after the
rebalance completes too:
https://github.com/apache/kafka/blob/trunk/clients/src/main/java/org/apache/kafka/clients/consumer/KafkaConsumer.java#L593
On Tue, Oct 6, 2015 at 5:27 PM, Gwen Shapira wrote:
> Ah,
Are you using the old or new producer? That sounds like the behavior the
old producer had -- it would stick to the same partition for awhile (10
minutes if I remember correctly). The new producer does not have this
behavior, preferring to round-robin the *available* brokers. Note that
since it
+1 non-binding. Verified artifacts, unit tests, quick start.
On Wed, Sep 9, 2015 at 10:09 AM, Guozhang Wang wrote:
> +1 binding, verified unit tests and quick start.
>
> On Wed, Sep 9, 2015 at 4:12 AM, Manikumar Reddy
> wrote:
>
> > +1 (non-binding).
Those APIs are not implemented in 0.8.2. They were included because the
APIs were being iterated on, but the implementation wasn't there yet.
You can expect to see those APIs (with some modifications as they've been
refined) in 0.8.3.
-Ewen
On Sun, Sep 6, 2015 at 10:42 AM, Phil Steitz
Steve,
I don't think there is a better solution at the moment. This is an easy
issue to miss in unit testing because generally connections to localhost
will be rejected immediately if there isn't anything listening on the port.
If you're running in an environment where this happens normally, then
Muqtafi,
There are corresponding Java or Scala classes. You can use them directly,
but beware that they are not considered public interfaces, so there are no
promises about compatibility. They could completely change between
releases. (The command line tools themselves, however, are considered
, or are batches just held in memory
before being sent to Kafka (or some other option)?
Thanks!
On Aug 26, 2015, at 9:50 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
Hemanth,
The Confluent Platform 1.0 version of have JSON embedded format support
(i.e. direct embedding of JSON messages
Hemanth,
Can you be a bit more specific about your setup? Do you have control over
the format of the request bodies that reach HAProxy or not? If you do,
Confluent's REST proxy should work fine and does not require the Schema
Registry. It supports both binary (encoded as base64 so it can be
schema registry ?
--regards
Hemanth
-Original Message-
From: Ewen Cheslack-Postava [mailto:e...@confluent.io]
Sent: Thursday, August 27, 2015 9:14 AM
To: users@kafka.apache.org
Subject: Re: Http Kafka producer
Hemanth,
Can you be a bit more specific about your setup? Do you have
initiated (and as long as
the dns entry is there) it only fails to connect in the poll() method. And
in the poll() method the status is not reset to DISCONNECTED and so it not
blacked out.
On Fri, Aug 21, 2015 at 10:06 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
Are you seeing
Are you seeing this in practice or is this just a concern about the way the
code currently works? If the broker is actually down and the host is
rejecting connections, the situation you describe shouldn't be a problem.
It's true that the NetworkClient chooses a fixed nodeIndexOffset, but the
There's not a precise date for the release, ~1.5 or 2 months from now.
On Fri, Aug 14, 2015 at 3:45 PM, Abhijith Prabhakar abhi.preda...@gmail.com
wrote:
Thanks Ewen. Any idea when we can expect 0.8.3?
On Aug 14, 2015, at 5:36 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
Hi
Hi Abhijith,
You should be using KafkaProducer, but KafkaConsumer is not ready yet. The
APIs are included in 0.8.2.1, but the implementation is not ready. Until
0.8.3 is released, you cannot rely only on kafka-clients if you want to
write a consumer. You'll need to depend on the main kafka jar
It's not public API so it may not be stable between releases, but you could
try using the ReassignPartitionsCommand class directly. Or, you can see
that the code in that class is a very simple use of ZkUtils, so you could
just make the necessary calls to ZkUtils directly.
In the future, when
You can use SimpleConsumer.getOffsetsBefore to get a list of offsets before
a Unix timestamp. However, this isn't per-message. The offests returned are
for the log segments stored on the broker, so the granularity will depend
on your log rolling settings.
-Ewen
On Wed, Aug 5, 2015 at 2:11 AM,
:52 PM, Joe Lawson
jlaw...@opensourceconnections.com wrote:
Ewen,
Do you have an example or link for the changes/plans that will bring the
benefits you describe?
Cheers,
Joe Lawson
On Aug 10, 2015 3:27 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
You can do this using
You can do this using the SimpleConsumer. See
https://cwiki.apache.org/confluence/display/KAFKA/0.8.0+SimpleConsumer+Example
for details with some code.
When the new consumer is released in 0.8.3, this will get a *lot* simpler.
-Ewen
On Fri, Aug 7, 2015 at 9:26 AM, Padgett, Ben
Kafka doesn't track per-message timestamps. The request you're using gets a
list of offsets for *log segments* with timestamps earlier than the one you
specify. If you start consuming from the offset returned, you should find
the timestamp you specified in the same log file.
-Ewen
On Mon, Aug
,
Nilesh Chhapru.
On Tuesday 28 July 2015 12:37 PM, Ewen Cheslack-Postava wrote:
Nilesh,
It's expected that a lot of memory is used for cache. This makes sense
because under the hood, Kafka mostly just reads and writes data to/from
files. While Kafka does manage some in-memory data, mostly
stops reading at times.
When i do a free -m on my broker node after 1/2 - 1 hr the memory foot
print is as follows.
1) Physical memory - 500 MB - 600 MB
2) Cache Memory - 6.5 GB
3) Free Memory - 50 - 60 MB
Regards,
Nilesh Chhapru.
On Monday 27 July 2015 11:02 PM, Ewen Cheslack-Postava
Hi Prabhjot,
Confluent has a REST proxy with docs that may give some guidance:
http://docs.confluent.io/1.0/kafka-rest/docs/intro.html The new producer
that it uses is very efficient, so you should be able to get pretty good
throughput. You take a bit of a hit due to the overhead of sending data
Having the OS cache the data in Kafka's log files is useful since it means
that data doesn't need to be read back from disk when consumed. This is
good for the latency and throughput of consumers. Usually this caching
works out pretty well, keeping the latest data from your topics in cache
and
log.segment.bytes? Thanks.
On Mon, Jul 27, 2015 at 1:25 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
I think log.cleanup.interval.mins was removed in the first 0.8 release.
It
sounds like you're looking at outdated docs. Search for
log.retention.check.interval.ms here:
http
log.retention.check.interval.ms, but there is
log.cleanup.interval.mins, is that what you mean?
If I set log.roll.ms or log.cleanup.interval.mins too small, will it hurt
the throughput? Thanks.
On Fri, Jul 24, 2015 at 11:03 PM, Ewen Cheslack-Postava e...@confluent.io
wrote:
You'll want to set the log retention
Try the --replica-assignment option for kafka-topics.sh. It allows you to
specify which brokers to assign as replicas instead of relying on the
assignments being made automatically.
-Ewen
On Mon, Jul 27, 2015 at 12:25 AM, Jilin Xie jilinxie1...@gmail.com wrote:
Hi
Is it possible to
You'll want to set the log retention policy via
log.retention.{ms,minutes,hours} or log.retention.bytes. If you want really
aggressive collection (e.g., on the order of seconds, as you specified),
you might also need to adjust log.segment.bytes/log.roll.{ms,hours} and
,daemonRegistryDir=/root/.gradle/daemon,pid=26478,idleTimeout=12,daemonOpts=-XX:MaxPermSize=512m,-Xmx1024m,-Dfile.encoding=UTF-8,-Duser.country=PH,-Duser.language=en,-Duser.variant]}.
04:32:54.865 [DEBUG] [org.gradle.launch
On Fri, Jul 24, 2015 at 3:34 AM, Ewen Cheslack-Postava e...@confluent.io
Also, the branch you're checking out is very old. If you want the most
recent release, that's tagged as 0.8.2.1. Otherwise, you'll want to use the
trunk branch.
-Ewen
On Thu, Jul 23, 2015 at 11:45 AM, Gwen Shapira gshap...@cloudera.com
wrote:
Sorry, we don't actually do SBT builds anymore.
This is a known issue. There are a few relevant JIRAs and a KIP:
https://issues.apache.org/jira/browse/KAFKA-1788
https://issues.apache.org/jira/browse/KAFKA-2120
https://cwiki.apache.org/confluence/display/KAFKA/KIP-19+-+Add+a+request+timeout+to+NetworkClient
-Ewen
On Tue, Jul 21, 2015 at 7:05
On Tue, Jul 21, 2015 at 2:38 AM, Stevo Slavić ssla...@gmail.com wrote:
Hello Apache Kafka community,
I find new consumer poll/seek javadoc a bit confusing. Just by reading docs
I'm not sure what the outcome will be, what is expected in following
scenario:
- kafkaConsumer is instantiated
Since you mentioned consumer groups, I'm assuming you're using the high
level consumer? Do you have auto.commit.enable set to true?
It sounds like when you start up you are always getting the
auto.offset.reset behavior, which indicates you don't have any offsets
committed. By default, that
,
Chandrash3khar Kotekar
Mobile - +91 8600011455
On Fri, Jul 17, 2015 at 4:57 AM, Ewen Cheslack-Postava e...@confluent.io
wrote:
Chandrashekhar,
If the firewall rules allow any TCP connection on those ports, you can
just
use Kafka directly and change the default port. If they actually verify
Chandrashekhar,
If the firewall rules allow any TCP connection on those ports, you can just
use Kafka directly and change the default port. If they actually verify
that its HTTP traffic then you'd have to the REST Proxy Edward mentioned or
another HTTP-based proxy.
-Ewen
On Thu, Jul 16, 2015 at
,
Thank you for your patient explaining. It is very helpful.
Can we assume that the long latency of ProducerPerformance comes from
queuing delay in the buffer and it is related to buffer size?
Thank you!
best,
Yuheng
On Thu, Jul 16, 2015 at 12:21 AM, Ewen Cheslack-Postava e...@confluent.io
Hi Jeff,
The simple consumer hasn't really changed, the info you found should still
be relevant. The wiki page on it might be the most useful reference for
getting started:
https://cwiki.apache.org/confluence/display/KAFKA/0.8.0+SimpleConsumer+Example
And if you want a version all setup to
Also worth mentioning is that the new producer doesn't have this behavior
-- it will round robin over available partitions for records without keys.
Available means it currently has a leader -- under normal cases this
means it distributes evenly across all partitions, but if a partition is
down
The tests are meant to evaluate different things and the way they send
messages is the source of the difference.
EndToEndLatency works with a single message at a time. It produces the
message then waits for the consumer to receive it. This approach guarantees
there is no delay due to queuing. The
at?
Thanks.
Yuheng
On Tue, Jul 14, 2015 at 12:18 AM, Ewen Cheslack-Postava
e...@confluent.io
wrote:
I implemented (nearly) the same basic set of tests in the system
test
framework we started at Confluent and that is going to move
Tim,
Kafka can be used as a key-value store if you turn on log compaction:
http://kafka.apache.org/documentation.html#compaction You need to be
careful with that since it's purely last-writer-wins and doesn't have
anything like CAS that might help you manage concurrent writers, but the
basic
I implemented (nearly) the same basic set of tests in the system test
framework we started at Confluent and that is going to move into Kafka --
see the wip patch for KIP-25 here: https://github.com/apache/kafka/pull/70
In particular, that test is implemented in benchmark_test.py:
The logic you're requesting is basically what the new producer implements.
The first condition is the batch size limit and the second is linger.ms.
The actual logic is a bit more complicated and has some caveats dealing
with, for example, backing off after failures, but you can see in this code
It does balance data, but is sticky over short periods of time (for some
definition of short...). See this FAQ for an explanation:
https://cwiki.apache.org/confluence/display/KAFKA/FAQ#FAQ-Whyisdatanotevenlydistributedamongpartitionswhenapartitioningkeyisnotspecified
?
This behavior has been
Current partition assignment only has a few limited options -- see the
partition.assignment.strategy consumer option (which seems to be listed
twice, see the second version for a more detailed explanation). There has
been some discussion of making assignment strategies user extensible to
support
It looks like you have mixed up versions of the kafka jars:
4. kafka_2.11-0.8.3-SNAPSHOT.jar
5. kafka_2.11-0.8.2.1.jar
6. kafka-clients-0.8.2.1.jar
I think org.apache.kafka.common.utils.Utils is very new, probably post
0.8.2.1, so it's probably caused by the kafka_2.11-0.8.3-SNAPSHOT.jar being
that could be rebuilt from the log
or a snapshot.
On lør. 13. jun. 2015 at 20.26 Ewen Cheslack-Postava e...@confluent.io
wrote:
Jay - I think you need broker support if you want CAS to work with
compacted topics. With the approach you described you can't turn on
compaction since that would make
:
But wouldn't the key-offset table be enough to accept or reject a write?
I'm not familiar with the exact implementation of Kafka, so I may be wrong.
On lør. 13. jun. 2015 at 21.05 Ewen Cheslack-Postava e...@confluent.io
wrote:
Daniel: By random read, I meant not reading the data
Jay - I think you need broker support if you want CAS to work with
compacted topics. With the approach you described you can't turn on
compaction since that would make it last-writer-wins, and using any
non-infinite retention policy would require some external process to
monitor keys that might
The new consumer implementation, which should be included in 0.8.3, only
needs a bootstrap.servers setting and does not use a zookeeper connection.
On Tue, Jun 9, 2015 at 1:26 PM, noah iamn...@gmail.com wrote:
We are setting up a new Kafka project (0.8.2.1) and are trying to go
straight to
It's not being switched in this case because the broker hasn't failed. It
can still connect to all the other brokers and zookeeper. The only failure
is of the link between a client and the broker.
Another way to think of this is to extend the scenario with more producers.
If I have 100 other
-Postava - do you have an example you could post?
From: Ewen Cheslack-Postava [e...@confluent.io]
Sent: Tuesday, May 19, 2015 3:12 PM
To: users@kafka.apache.org
Subject: Re: KafkaConsumer poll always returns null
The new consumer in trunk is functional
The new consumer in trunk is functional when used similarly to the old
SimpleConsumer, but none of the functionality corresponding to the high
level consumer is there yet (broker-based coordination for consumer
groups). There's not a specific timeline for the next release (i.e. when
it's ready).
201 - 300 of 340 matches
Mail list logo