Hi,
I am not sure if I understand correctly: If you use default
TimestampExtractor, the whole pipeline will be event-time based.
However, as you want to compute the AVG, I would recommend a different
pattern anyway:
FEED -> groupByKey() -> window() -> aggregate() -> mapValues() = avgKTable
In
Yes. Key can be of any type and we cannot enforce immutable types at API
level, and thus, it could get modified as a "side effect".
The problem is, that if the key would be modified, it would corrupt data
partitioning and thus would lead to wrong result.
It's not possible to modify the key via re
I think I have an understanding of how Kafka Streams is handling time
behind the scenes and would like someone to verify it for me. The actual
reason is I am running into behavior where I only can join two streams for
a little, then it stops working.
Assuming a topology like this:
FEED -> groupB
This is a trace-level log which means that consumer already creates a fetch
request to the given node from which you are reading data so no more requests
cannot be created. Did you get any other warn-level or error-level logs when
failing to fetching message?
发
Yes, I have upgraded my cluster and client both to version 10.2.1 and
currently monitoring the situation.
Will report back in case I find any errors. Thanks for the help though.
-Sameer.
On Fri, May 5, 2017 at 3:37 AM, Matthias J. Sax
wrote:
> Did you see Eno's reply?
>
> Please try out Streams
Hello,
I am using kafka 0.10.1.0 and failing to fetch messages with following error
message in the log :
Skipping fetch for partition
MYPARTITION because there is an in-flight request to MYMACHINE:9092 (id: 0
rack: null) (org.apache.kafka.clients.consumer.
So the reasoning is that the key may be a mutable object which in term
could potentially cause disaster?
Just clarifying because I think the initializer should only return a value
(as it does right now).
On Thu, May 4, 2017 at 3:02 PM Matthias J. Sax
wrote:
> Currently, you don't get any hold on
If your consumer fails (ie. whole process dies) setting the value high
is not a problem (because the heartbeat thread dies, too, and the
failure will be detected quickly).
It's only a problem if you "main processing thread" dies (and everything
else is still up and running), or if you main process
Hi,
we don't believe in triggers ;)
https://www.confluent.io/blog/watermarks-tables-event-time-dataflow-model/
-> Thus, it's a BigQuery flaw to not support updates... (IMHO)
(We are also considering improving KStream-KStream join though, but
that's of course no short term solution for you:
https
About
> 07:44:08.493 [StreamThread-10] INFO o.a.k.c.c.i.AbstractCoordinator -
> Discovered coordinator broker-05:6667 for group group-2.
Please upgrade to Streams 0.10.2.1 -- we fixed couple of bug and I would
assume this issue is fixed, too. If not, please report back.
> Another question that I
The timestamp is always stored -- it's a metadata field that was added
to the message format in 0.10.0.
By default, Producer uses System.currentTimeMillis() to set the
timestamp before it send the record to the broker. Or you can explicitly
set the timestamp by yourself.
The default TimestampExtr
>
>
> The documentation says "Upgrading from 0.8.x or 0.9.x to 0.10.0.0"
>
> I am upgrading from kafka_2.9.2-0.8.1.1 so which one is correct
>
> A. 0.8.1.1
>
> *inter.broker.protocol.version**=**0.8.1.1*
>
> *log.message.format.version**=**0.8.1.1*
>
> B. 0.8.1
>
> *inter.broker.protocol.version**=
Did you see Eno's reply?
Please try out Streams 0.10.2.1 -- this should be fixed there. If not,
please report back.
I would also recommend to subscribe to the list. It's self-service
http://kafka.apache.org/contact
-Matthias
On 5/3/17 10:49 PM, Sameer Kumar wrote:
> My brokers are on version 1
Currently, you don't get any hold on the key, because the key must be
protected from modification.
Thus, the initial value must be the same for all keys atm.
We are aware that this limits some use cases and there is already a
proposal to add keys to some API.
https://cwiki.apache.org/confluence/d
Upgrading from kafka_2.9.2-0.8.1.1 to kafka_2.11-0.10.0.0
The new version kafka will look at the same location for log files as older
one is what I am assuming.
As per documentation following properties will be set in the new broker
inter.broker.protocol.version=0.8.1.1
log.message.format.versi
Hi,
In Producer record, we have timestamp.
- Where it is actually stored, if stored?
- When producing resorts with KStream.to() which timestamp is used, if used?
In order to guarantee proper timestamp, do we have to always implement
timestamp extractor?
Thanks in advance.
Hi Matthias,
Please find the answers below.
I would recommend to double check the following:
>
> - can you confirm that the filter does not remove all data for those
> time periods?
>
Filter does not remove all data. There is a lot of data coming in even
after the filter stage.
> - I would a
Hi Guys,
I found one of topic's time index size is zero, we are using kafka version
0.10.1.1. is there anyone have any idea why the zero time index will be
happened ?
thanks.
the following is the data snippet:
-rw-r--r-- 1 root root0 May 4 05:36
/kafka/data/thl_raw-1/000114296521.t
Hi guys,
I want to perform a join between two KStreams.
An event may appear only on one of the streams (either one of them), so I
can't use inner join (which emits only on a match) or left join (which
emits only when the left input arrives).
This leaves me with outer join. The problem with outer j
What will be the impact of setting the consumer config max.poll.interval.ms
to 10 or 15 mins?
How will impact nodes joining , nodes leaving, nodes crashing.?
Chaitanya
--
Chaitanya
Thanks Eno,
We still see problems on our side.
when we run kafka-streams 0.10.1.1 eventually the problem goes away but
with 0.10.2.1 it is not.
We see a lot of the rebalancing messages I wrote before
on at least 1 kafka-stream nodes we see disconnection messages like the
following. These messages
kafka-python, yes.
On May 4, 2017 2:28 AM, "Paul van der Linden" wrote:
Thanks everyone. @Dana is that using the kafka-python library?
On Thu, May 4, 2017 at 4:52 AM, Dana Powers wrote:
> Requires stopping your existing consumers, but otherwise should work:
>
> from kafka import KafkaConsumer
-- Forwarded message --
From: Abhimanyu Nagrath
Date: Thu, May 4, 2017 at 4:21 PM
Subject: Error: Executing consumer group command failed due to Request
GROUP_COORDINATOR failed on brokers List(localhost:9092 (id: -1 rack:
null))
To: d...@kafka.apache.org
Hi, I am using single no
Thanks everyone. @Dana is that using the kafka-python library?
On Thu, May 4, 2017 at 4:52 AM, Dana Powers wrote:
> Requires stopping your existing consumers, but otherwise should work:
>
> from kafka import KafkaConsumer, TopicPartition, OffsetAndMetadata
>
> def reset_offsets(group_id, topic,
It is odd as the person that originally reported the problem has verified
that it is fixed.
On Thu, 4 May 2017 at 08:31 Guozhang Wang wrote:
> Ara,
>
> That is a bit weird, I double checked and agreed with Eno that this commit
> is in both trunk and 0.10.2, so I suspect the same issue still pers
Ara,
That is a bit weird, I double checked and agreed with Eno that this commit
is in both trunk and 0.10.2, so I suspect the same issue still persists in
trunk, hence there might be another issue that is not fixed in 2645. Could
you help verify if that is the case? In which we can re-open
https:/
26 matches
Mail list logo