Message loss in kafka when using java API.

2012-11-29 Thread boris kogan
Hi, All We encounter an issue with Kafka of messages loss: We have Kafka 0.7.0 installed in cluster mode on three nodes(on partition per node configured). We are using kafka java API to send bulk of messages(1000 messages in each) in sync mode. After the last bulk is sent we close the producer.

Re: Unexpected end of ZLIB input stream

2012-11-29 Thread Dmitri Priimak
On 12-11-28 09:54 PM, Jun Rao wrote: Dmitri, Could you reproduce this easily? Are you using a load balancer? Earlier, another user had the same issue and eventually figured out that the problem is in the network router. I do not believe I have any loadbalancer anywhere in the picture and by the

RE: producer queue size

2012-11-29 Thread Jamie Wang
log.flush.interval is set to 1. Should I increase or decrease? thanks Jamie -Original Message- From: Jun Rao [mailto:jun...@gmail.com] Sent: Wednesday, November 28, 2012 10:00 PM To: users@kafka.apache.org Subject: Re: producer queue size Is the server busy on I/O? What's log.flush

Re: Message loss in kafka when using java API.

2012-11-29 Thread Neha Narkhede
Hi Boris, In Kafka 0.7, the producer does not get any ACK from the server, so the protocol is fire and forget. What could happen is if a broker is shutting down, the messages might still live in the producer's socket buffer and the server can shutdown before getting a chance to read the producer's

Re: consumer read process

2012-11-29 Thread S Ahmed
testing new list On Wed, Nov 28, 2012 at 10:44 AM, Jun Rao wrote: > You can find the information at > http://incubator.apache.org/kafka/design.html > > Look for consumer registration algorithm and consumer rebalancing > algorithm. > > Thanks, > > Jun > > > > On Wed, Nov 28, 2012 at 7:13 AM, S A

FW: Zookeeper Configuration Question

2012-11-29 Thread Sybrandy, Casey
Re-sending since I didn't get any responses before and want to make sure this made it to the correct group. From: Sybrandy, Casey Sent: Wednesday, November 28, 2012 2:39 PM To: users@kafka.apache.org Subject: Zookeeper Configuration Question All, We're having an

Consumer Hanging (yes, read the FAQ)

2012-11-29 Thread Juan Valencia
My consumers randomly hang. They essentially end up waiting on a kafka stream: There is plenty of data on the brokers and a ratio of about 1 consumer for 5 partitions. Any thoughts on how to keep the stream going? Or at least restart on hang? e.g. ConsumerConfig consumerConfig =

Fwd: Consumer Hanging (yes, read the FAQ)

2012-11-29 Thread Juan Valencia
-- Forwarded message -- From: Juan Valencia Date: Thu, Nov 29, 2012 at 11:45 AM Subject: Consumer Hanging (yes, read the FAQ) To: users@kafka.apache.org My consumers randomly hang. They essentially end up waiting on a kafka stream: There is plenty of data on the brokers and a r

Re: Consumer Hanging (yes, read the FAQ)

2012-11-29 Thread Matan Amir
A popular case is that the thread isn't catching thrown Exceptions in run() - which causes the thread to be restarted in a typical Executor configuration - and does not pull any more items. Is that the case? On Thu, Nov 29, 2012 at 11:48 AM, Juan Valencia wrote: > -- Forwarded message -

Re: FW: Zookeeper Configuration Question

2012-11-29 Thread James A. Robinson
On Thu, Nov 29, 2012 at 10:46 AM, Sybrandy, Casey wrote: > Re-sending since I didn't get any responses before and want to make > sure this made it to the correct group. >>We're having an issue with Zookeeper, which has nothing to do with >>Kafka, but my consumers don't appear to be attempting to

Re: FW: Zookeeper Configuration Question

2012-11-29 Thread James A. Robinson
On Thu, Nov 29, 2012 at 1:15 PM, James A. Robinson wrote: > For my kafka startup I point to the zookeeper "cluster" like so: > > --kafka-zk-connect > logproc-dev-03:2181,logproc-dev-03:2182,logproc-dev-03:2183 Sorry, wrong copy and paste! For the kafka startup I point to the zookeeper cluster

Re: Consumer Hanging (yes, read the FAQ)

2012-11-29 Thread Neha Narkhede
The thread dump basically shows that the consumer thread (from your executor) is waiting on the internal fetcher queue. This means the internal fetcher queue is empty, which in turn suggests that the fetcher might not be running. Please can you file a bug and share the entire thread dump there ? T

Re: FW: Zookeeper Configuration Question

2012-11-29 Thread Neha Narkhede
Please can you send around the log that shows the zookeeper connection error ? I would like to see if it fails at connection establishment or session establishment. Thanks, Neha On Thu, Nov 29, 2012 at 1:19 PM, James A. Robinson wrote: > On Thu, Nov 29, 2012 at 1:15 PM, James A. Robinson > wrot

Re: Unexpected end of ZLIB input stream

2012-11-29 Thread Neha Narkhede
Dmitri, You said that the problem goes away once you upgrade to 0.7.1. Do you mean that the message corruption doesn't happen or that when it happens, the broker doesn't shut down ? I'm asking since we occasionally do see the corruption on 0.7.1 but don't have a good way to reproduce that. 0.7.1 f

Re: Unexpected end of ZLIB input stream

2012-11-29 Thread Dmitri Priimak
On 11/29/2012 01:50 PM, Neha Narkhede wrote: > Dmitri, > > You said that the problem goes away once you upgrade to 0.7.1. Do you > mean that the message corruption doesn't happen or that when it > happens, the broker doesn't shut down ? I do not see corrupt message at all. > I'm asking since we o

Re: Unexpected end of ZLIB input stream

2012-11-29 Thread Neha Narkhede
That just meant that we couldn't reproduce it during our testing, but we occasionally do see it in production, which is much harder to reproduce. It will be great if you can reproduce the issue and attach the test case. Thanks, Neha On Thu, Nov 29, 2012 at 2:03 PM, Dmitri Priimak wrote: > On 11/

Re: Unexpected end of ZLIB input stream

2012-11-29 Thread Dmitri Priimak
On 11/29/2012 02:06 PM, Neha Narkhede wrote: > That just meant that we couldn't reproduce it during our testing, but > we occasionally do see it in production, which is much harder to > reproduce. > It will be great if you can reproduce the issue and attach the test case. I see. I will try to isola

Java API documentation

2012-11-29 Thread Subhash Agrawal
Hi All, I am new to Kafka and we are planning to use Kafka as message broker. I am wondering if there are java API documentation for Kafka. I noticed that current API doc is written for Scala user. I believe it could be because it is written in Scala. Please let me know. Thanks Subhash A.

Re: Java API documentation

2012-11-29 Thread Joel Koshy
Are you referring to the scaladoc? http://people.apache.org/~joestein/kafka-0.7.1-incubating-docs/ There's (some) documentation on the java api as well (i.e., the javaapi packages) that you can use from Java. Joel On Thu, Nov 29, 2012 at 3:14 PM, Subhash Agrawal wrote: > Hi All, > > I am new to

Re: Java API documentation

2012-11-29 Thread Jay Kreps
I think what Subhash is asking for is just plain old fashioned javadoc. People are used to having that for java apis. We should do that, but I think it would require moving the java-friendly wrappers into java. This would probably be a good thing to do anyway just to get rid of all the crazy scala

Re: producer queue size

2012-11-29 Thread Jun Rao
Hmm, that should be big enough. Thanks, Jun On Thu, Nov 29, 2012 at 10:03 AM, Jamie Wang wrote: > log.flush.interval is set to 1. Should I increase or decrease? > > thanks > Jamie > > -Original Message- > From: Jun Rao [mailto:jun...@gmail.com] > Sent: Wednesday, November 28, 2012