Hi Manas,
you need to make sure to differentiate between what Flink calls
"pre-flight phase" and "cluster phase".
The pre-flight phase is were the pipeline is constructed and all
functions are instantiated. They are then later serialized and send to
the cluster.
If you are reading your pro
Hi Jon,
I would not recommend to use the configuration parameter. It is not
deprecated yet but can be considered legacy code from before we reworked
the type system.
Regards,
Timo
On 16.10.20 13:23, Kurt Young wrote:
Yes, I think this is a bug, feel free to open a jira and a pull request.
uplicates, but for records with no duplicates, I'd have to wait until
no more records are coming -- am I missing something?
Thanks so much,
Austin
On Fri, Oct 9, 2020 at 10:44 AM Timo Walther <mailto:twal...@apache.org>> wrote:
Hi Austin,
if you don't want to worry a
Hi Rex,
let me copy paste my answer from a similar thread 2 months ago:
Hi,
this might be helpful as well:
https://lists.apache.org/thread.html/rfe3b45a10fc58cf19d2f71c6841515eb7175ba731d5055b06f236b3f%40%3Cuser.flink.apache.org%3E
First of all, it is important to know if you are interested i
Hi Luan,
we haven't updated all config parameters to string-based options. This
is still on going. The idle state retention will be configurable in 1.12:
https://issues.apache.org/jira/browse/FLINK-18555
I hope this helps.
Regards,
Timo
On 09.10.20 15:33, Luan Cooper wrote:
Hi
I've read
ub.com/austince/flink-1.10-sql-windowing-error
On Mon, Oct 5, 2020 at 4:24 AM Timo Walther <mailto:twal...@apache.org>> wrote:
Btw which planner are you using?
Regards,
Timo
On 05.10.20 10:23, Timo Walther wrote:
> Hi Austin,
>
> could you share
28, 2020 at 6:41 AM Timo Walther <mailto:twal...@apache.org>> wrote:
Hi Dan,
unfortunetely, it is very difficult to read you plan? Maybe you can
share a higher resolution and highlight which part of the pipeline
is A,
B etc. In general, the planner should be smart
Btw which planner are you using?
Regards,
Timo
On 05.10.20 10:23, Timo Walther wrote:
Hi Austin,
could you share some details of your SQL query with us? The reason why
I'm asking is because I guess that the rowtime field is not inserted
into the `StreamRecord` of DataStream API. The ro
es, it should also work for ingestion time.
I am not entirely sure whether event time is preserved when
converting a Table into a retract stream. It should be possible
and if it is not working, then I guess it is a missing feature.
But I am sure that @Timo Walther
Hi Sunitha,
currently, not every connector can be mixed with every API. I agree that
it is confusing from time to time. The HBase connector is an
InputFormat. DataSet, DataStream and Table API can work with
InputFormats. The current Hbase input format might work best with Table
API. If you li
Hi Dan,
unfortunetely, it is very difficult to read you plan? Maybe you can
share a higher resolution and highlight which part of the pipeline is A,
B etc. In general, the planner should be smart enough to reuse subplans
where appropriate. Maybe this is a bug or shortcoming in the optimizer
r
Hi Dan,
could you share the plan with us using `TableEnvironment.explainSql()`
for both queries?
In general, views should not have an impact on the performance. They are
a logical concept that gives a bunch of operations a name. The contained
operations are inlined into the bigger query duri
Hi Piyush,
unfortunately, UDFs have no direct access to Flink's state. Aggregate
functions are the only type of functions that can be stateful at the
moment. Aggregate functions store their state in an accumulator that is
serialized/deserialized on access, but an accumulator field can be
back
Hi Husky,
I guess https://issues.apache.org/jira/browse/FLINK-14055 is what is
needed to make this feature possible.
@Rui: Do you know more about this issue and current limitations.
Regards,
Timo
On 18.09.20 09:11, Husky Zeng wrote:
When we submit a job which use udf of hive , the job will
Hi Lian,
you are right that timers are not available in a ProcessWindowFunction
but the state store can be accessed. So given that your window width is
1 min, you could maintain an additional state value for counting the
minutes and updating your counter once this value reached 60.
Otherwise
Hi Arpith,
is there a JIRA ticket for this issue already? If not, it would be great
if you can report it. This sounds like a critical priority issue to me.
Thanks,
Timo
On 22.09.20 06:25, Arpith P wrote:
Hi Peter,
I have recently had a similar issue where I could not load from the
checkpoi
Hi Luan,
this sound more of a new feature request to me. Maybe you can already
open an issue for it.
I will loop in Chesnay in CC if there is some possibility to achieve
this already?
Regards,
Timo
On 21.09.20 06:37, Luan Cooper wrote:
Hi
We're running flink standalone cluster on k8s
whe
Hi Saksham,
if I understand you correctly, you are running Zookeeper and Flink
locally on your machine? Are you using Docker or is this a bare metal
setup? The exception indicates that your paths contain `hdfs:` as URL
scheme. Are you sure you want to use HDFS? If yes, you need to add an
HDFS
Hi Ori,
first of all, watermarks are sent to all side outputs (this is tested
here [1]). Thus, operators in the side output branch of the pipeline
will work similar to operators in the main branch.
When calling `assignTimestampsAndWatermarks`, the inserted operator will
erase incoming waterm
Hi Satyam,
this has historical reasons. In the beginning all SQL queries were
embedded in Java programs and thus Java strings. So single quote was
handy for declaring SQL strings in a Java string and backticks for
escaping keywords. But I agree that we should make this configurable.
Feel free
Hi Violeta,
I just noticed that the plan might be generated from Flink's old planner
instead of the new, more performant Blink planner. Which planner are you
currently using?
Regards,
Timo
On 08.09.20 17:51, Timo Walther wrote:
You are using the old connectors. The new connector
r way?
Regards,
Heidy
----
*From:* Timo Walther mailto:twal...@apache.org>>
*Sent:* Wednesday, September 9, 2020 1:58 PM
*To:* user@flink.apache.org <mailto:user@flink.apache.org>
Hi Hazem,
I guess your performance is mostly driven by the serialization overhead
in this case. How do you declare your state type?
Flink comes with different serializers. Not all of them are extracted
automatically when using reflective extraction methods:
- Note that `Serializable` decla
ossible
sample source code for reference to stream database. Please help me
badly stuck.
In the mail, I see you asked me to register. Are you referring to any
training here or any other registration.
Regards,
Sunitha.
On Tuesday, September 8, 2020, 08:19:49 PM GMT+5:30, Timo Walther
wrote
You are using the old connectors. The new connectors are available via
SQL DDL (and execute_sql() API) like documented here:
https://ci.apache.org/projects/flink/flink-docs-release-1.11/dev/table/connectors/filesystem.html
Maybe this will give your some performance boost, but certainly not
eno
Hi Sunitha,
what you are describing is a typical streaming enrichment. We need to
enrich the stream with some data from a database. There are different
strategies to handle this:
1) You are querying the database for every record. This is usually not
what you want because it would slow down y
Hi Vijay,
one comment to add is that the performance might suffer with multiple
map() calls. For safety reason, records between chained operators are
serialized and deserialized in order to strictly don't influence each
other. If all functions of a pipeline are guaranteed to not modify
incomi
Hi Violeta,
can you share your connector code with us? The plan looks quite
complicated given the relatively simple query. Maybe there is some
optimization potential. But before we dive deeper, I see a `Map(to:
Row)` which indicates that we might work with a legacy sink connector.
Did you tr
m
On Tue, Sep 1, 2020 at 4:46 AM Timo Walther <mailto:t...@ververica.com>> wrote:
Hi Satyam,
Matthias is right. A rowtime attribute cannot be modified and needs
to be passed "as is" through the pipeline. The only exceptions are
if a newer rowtime is offered
Hi,
this might be helpful as well:
https://lists.apache.org/thread.html/rfe3b45a10fc58cf19d2f71c6841515eb7175ba731d5055b06f236b3f%40%3Cuser.flink.apache.org%3E
First of all, it is important to know if you are interested in
end-to-end tests (incl. connectors) or excluding connectors. If you jus
Hi Forideal,
luckily these problems will belong to the past in Flink 1.12 when UDAF
are updated to the new type system [1]. Lists will be natively supported
and registering custom KryoSerializers consistently as well.
Until then, another workaround is to override getAccumulatorType() and
def
Hi Lei,
you can check how the FlinkFixedPartitioner [1] or
Tuple2FlinkPartitioner [2] are implemented. Since you are using SQL
connectors of the newest generation, you should receive an instance of
org.apache.flink.table.data.RowData in your partitioner.
You can create a Maven project with a
Hi Lu,
`env.execute("table api");` is not necessary after FLIP-84 [1]. Every
method that has `execute` in its name will immediately execute a job.
Therefore your `env.execute` has an empty pipeline.
Regards,
Timo
[1]
https://wiki.apache.org/confluence/pages/viewpage.action?pageId=134745878
What you can do is creating an initial control stream e.g. using
`StreamExecutionEnivronment.fromElements()` and either use
`union(controlStream, actualStream)` or use
`actualStream.connect(controlStream)`.
Regards,
Timo
On 12.08.20 18:15, Andrey Zagrebin wrote:
I do not think so. Each timer
Hi Faye,
the problem lies in the wrong design of JDK's java.sql.Timestamp. You
can also find a nice summary in the answer here [1]. java.sql.Timestamp
is timezone dependent. Internally, we subtract/normalize the timezone
and work with the UNIX timestamp. Beginning from Flink 1.9 we are using
Hi Vincent,
we don't have a step by step guide for adding new operators. Most of the
important operations are exposed via DataStream API. Esp.
ProcessFunction [1] fits for most complex use cases with access to the
primitives such as time and state.
What kind of operator is missing for your u
is produces the expected output.
Also, I will assume that this is the best way to solve my problem - I
can't use Flink's session windows. Let me know if anyone has any other
ideas though!
Thank you for your time and quick response!
On Tue, Aug 11, 2020 at 1:45 PM Timo Walther <ma
Hi Faye,
Flink does not officially provide testing tools at the moment. However,
you can use internal Flink tools if they solve your problem.
The `flink-end-to-end-tests` module [1] shows some examples how we test
Flink together with other systems. Many tests are still using plain bash
scrip
Hi Dan,
InputFormats are the connectors of the DataSet API. Yes, you can use
either readFile, readCsvFile, readFileOfPrimitives etc. However, I would
recommend to also give Table API a try. The unified TableEnvironment is
able to perform batch processing and is integrated with a bunch of
conn
Hi Manas,
at the first glance your code looks correct to me. I would investigate
if your keys and watermarks are correct. Esp. the watermark frequency
could be an issue. If watermarks are generated at the same time as the
heartbeats itself, it might be the case that the timers fire first
befo
Hi Dongwon,
another possibility is to use DataStream API before. There you can
extract the metadata and use DataStream.assignTimestampsAndWatermarks
before converting the stream to a table.
Regards,
Timo
On 11.08.20 09:41, Dongwon Kim wrote:
Hi Dawid,
I'll try your suggestion [2] and wait
s about TableFunction, so maybe it is something different, but
related.
I have created a small github project with both cases:
https://github.com/dmytroDragan/flink-1.11-sql-agg-issue/blob/master/src/test/java/lets/test/flink/AggFunTest.java
I would appreciate if you could take a look.
On 27/07/2
recommended Maven version for building Flink.
@Felipe Can you provide us the full stacktrace? This could be a library
issue in regards to JDK compatibility.
On 27/07/2020 15:23, Timo Walther wrote:
Hi Felipe,
are you sure that Maven and the TaskManagers are using the JDK version
that you ment
Hi Dmytro,
aggregate functions will support the new type system in Flink 1.12.
Until then, they cannot be used with the new `call()` syntax as
anonymous functions. In order to use the old type system, you need to
register the function explicilty using SQL `CREATE FUNCTION a AS
'myFunc'` and t
Hi Felipe,
are you sure that Maven and the TaskManagers are using the JDK version
that you mentioned?
Usually, a `mvn clean install` in the `.../flink/` directory should
succeed without any problems. Also your Maven version seems pretty old.
I'm using Apache Maven 3.6.3 for example.
The No
you check that?
@Timo I think we should define the serialVersionUID for all the classes
which implements Serializable. What do you think?
Regards,
Dian
在 2020年7月27日,下午4:38,Timo Walther <mailto:twal...@apache.org>> 写道:
Hi,
the InvalidClassException indicates that you are using
them, but I`m wonder if
there is a better way to do it.
It sounds a quite strange that with having Blink planner which optimise
DataStream pipelines for stream and batch jobs,
there is necessity to write the same things on DataStream and DataSet API.
On 24/07/2020, 15:36, "Timo
Hi Aviral,
as far as I know we are not calling RocksDB API to perform snapshots. As
the Stackoverflow answer also indicates most of the snapshotting is done
outside of RocksDB by just dealing with the SST files. Have you checked
the available metrics in the web UI?
https://ci.apache.org/proj
Hi,
the InvalidClassException indicates that you are using different
versions of the same class. Are you sure you are using the same Flink
minor version (including the Scala suffix) for all dependencies and
Kubernetes?
Regards,
Timo
On 27.07.20 09:51, Wojciech Korczyński wrote:
Hi,
when
Hi Dmytro,
`StreamTableEnvironment` does not support batch mode currently. Only
`TableEnvironment` supports the unified story. I saw that you disabled
the check in the `create()` method. This check exists for a reason.
For batch execution, the planner sets specific properties on the stream
g
definitely the `AvroSerializer` if the type
information is `AvroTypeInfo`. You can check that via `dataStream.getType`.
I hope this helps.
Regards,
Timo
On 16.07.20 14:28, Timo Walther wrote:
Hi Lasse,
are you using Avro specific records? A look into the code shows that the
warnings in the
Hi Lasse,
are you using Avro specific records? A look into the code shows that the
warnings in the log are generated after the Avro check:
https://github.com/apache/flink/blob/master/flink-core/src/main/java/org/apache/flink/api/java/typeutils/TypeExtractor.java#L1741
Somehow your Avro object
Hi YI,
not all conversion might be supported in the `toRetractStream` method.
Unfortunately, the rework of the type system is still in progress. I
hope we can improve the user experience there quite soon.
Have you tried to use `Row` instead? `toRetractStream[Row]` should work
for all data ty
Hi Satyam,
I'm not aware of an API to solve all your problems at once. A common
pattern for failures in user-code is to catch errors in user-code and
define a side output for an operator to pipe the errors to dedicated
sinks. However, such a functionality does not exist in SQL yet. For the
si
Hi Lorenzo,
as far as I know we don't support Avro's logical times in Flink's
AvroInputFormat yet. E.g. AvroRowDeserializationSchema [1] supports the
1.8.2 version of logical types but might be incompatible with 1.9.2.
Reg 2) Specific record generated with AVRO 1.9.2 plugin:
Could you send u
Hi Annemarie,
if TTL is what you are looking for and queryable state is what limits
you, it might make sense to come up with a custom implementation of
queryable state? TTL might be more difficult to implement. As far as I
know this feature is more of an experimental feature without any
consi
Hi Martin,
usually, this error occurs when people forget to add
`org.apache.flink.api.scala._` to their imports. It is triggered by the
Scala macro that the DataStream API uses for extracting types.
Can you try to call `result.toAppendStream[Row]` directly? This should
work if you import `or
Hi Nikola,
the reason for deprecating `registerTableSource` is that we aim to have
everything declarative in Table API. A table program should simply
declare what it needs and the planner should find a suitable connector,
regardless how the underlying class structure looks like. This might
al
Hi Yuval,
currently there is no API for getting those insights. I guess you need
to use internal API for getting this information. Which planner and
version are you using?
Regards,
Timo
On 18.05.20 14:16, Yuval Itzchakov wrote:
Hi,
Is there any way to infer if a Table is going to generate
of the box, is it possible to add some extra operator
after sink which will always have watermark which is greater than
sink function watermarks,
as its a downstream operator.
Also, does the problem simplify if we have Kafka sink?
On Tue, Apr 28, 2020 at 10:35 PM Timo Walther mail
Hi Max,
as far as I know a better ML story for Flink is in the making. I will
loop in Becket in CC who may give you more information.
Regards,
Timo
On 28.04.20 07:20, m@xi wrote:
Hello Flinkers,
I am building a *streaming* prototype system on top of Flink and I want
ideally to enable ML tra
Hi Shubham,
you can call stream.process(...). The context of ProcessFunction gives
you access to TimerService which let's you access the current watermark.
I'm assuming your are using the Table API? As far as I remember,
watermark are travelling through the stream even if there is no
time-ba
Timo,
I am trying to convert simply back to a DataStream. Let's say:
DataStream>
I can convert the DataStream into a table without a problem, the problem
is getting a DataStream back.
Thanks
Gyula
On Tue, Apr 28, 2020 at 6:32 PM Timo Walther <mailto:twal...@apache.org>> wrot
Hi Manas,
Reg. 1: I would recommend to use a debugger in your IDE and check which
watermarks are travelling through your operators.
Reg. 2: All event-time operations are only performed once the watermark
arrived from all parallel instances. So roughly speaking, in machine
time you can assume
Hi Gyula,
are you coming from DataStream API or are you trying to implement a
source/sink? It looks like the array is currently serialized with Kryo.
I would recommend to take a look at this class:
org.apache.flink.table.types.utils.LegacyTypeInfoDataTypeConverter
This is the current mapping
Hi Gyula,
first of all the exception
```
org.apache.flink.table.api.TableException: Rowtime attributes must not
be in the input rows of a regular join. As a workaround you can cast the
time attributes of input tables to TIMESTAMP before.
```
is IMHO one of the biggest shortcomings that we cu
Yes. But that's the problem of your use cases, right? If you need to
wait for the sink to be completed, it is not a terminating operator anymore.
Regards,
Timo
On 15.04.20 10:50, KristoffSC wrote:
Thank you very much for your answer.
I have a question regarding your first paragraph:
" it req
Hi Dominik,
Flink does not use Java serialization logic for network communication.
So objects must not implement `Serializable` for usage during runtime
(DataStream).
Only if those classes are member variables of a Function like
MapFunction, they need to serializable to ship the function cod
Hi Kristoff,
synchronization across operators is not easy to achieve.
If one needs to wait until a sink has processed some element, it
requires that a sink participates in the pipeline. So it is not located
as a "leaf" operator but location somewhere in the middle.
So your idea to call MQ di
Hi Dmytro,
table function will be supported in Flink 1.11 with the new type system.
Hopefully, we can also support aggregate functions until then.
Regards,
Timo
On 14.04.20 15:33, godfrey he wrote:
Hi Dmytro,
Currently, TableEnvironment does not support
register AggregationFunction and Tab
Hi Oleg,
this sounds indeed like abnormal behavior. Are you sure that these large
checkpoints are related to the Kafka consumer only? Are there other
operators in the pipeline? Because internally the state kept in a Kafka
consumer is pretty minimal and only related to Kafka partition and
offs
Hi,
currently we don't provide more flexible windowing semantics in SQL. For
this, a programmatic API like the DataStream API is a better fit with
custom triggers and other more advanced features.
Regards,
Timo
On 14.04.20 13:31, snack white wrote:
Hi,
In flink sql session window, is ther
Hi Navneeth,
it might be also worth to look into Ververica Plaform for this. The
community edition was published recently is free of charge. It provides
first class K8s support [1].
There is also a tutorial how to deploy it on EKS [2] (not the most
recent one through).
Regards,
Timo
[1]
Hi Gyula,
some disclaimer: the type system rework is still ongoing and there a a
couple of known issues and missing end-to-end tests around this topic.
I would therefore recommend to declare the sink as `STRING NULL` for now.
Can you open an issue for your concrete use case with some example
Hi Jark,
thanks for sharing the results. I think for the databases usage part,
I'm pretty sure that we could also rely on some Gartner landscape
research. Just an idea.
Thanks for performing the survey!
Timo
On 03.04.20 10:17, Jark Wu wrote:
Hi everyone,
Thanks all for the feedbacks! I wo
Hi Kristoff,
case 1:
first of all Flink groups keys internally into so-called "key groups"
for reducing the management overhead. The maximum parallelism decides
about the number of key groups. When performing a rescale, the key
groups are basically distributed using some consistent hashing al
Hi Manas,
first of all, after assigning watermarks at the source level, usually
Flink operators make sure to handle the watermarks.
In case of a `union()`, the subsequent operator will increment its
internal event-time clock and emit a new watermark only if all input
streams (and their paral
Hi Robin,
this is a very good observation and maybe even unintended behavior.
Maybe Arvid in CC is more familiar with the checkpointing?
Regards,
Timo
On 02.04.20 15:37, Robin Cassan wrote:
Hi all,
I am wondering if there is a way to make a flink job fail (not cancel
it) when one or sever
Or better: "But for sources, you need to emit a watermark from all
sources in order to have progress in event-time."
On 24.03.20 13:09, Timo Walther wrote:
Hi,
1) yes with "partition" I meant "parallel instance".
If the watermarking is correct in the DataStre
watermark progresses
and the output is generated?
Best Regards,
Dom.
wt., 24 mar 2020 o 10:01 Timo Walther <mailto:twal...@apache.org>> napisał(a):
Hi Dominik,
the big conceptual difference between DataStream and Table API is that
record timestamps are part of the schema in
Hi Dominik,
the big conceptual difference between DataStream and Table API is that
record timestamps are part of the schema in Table API whereas they are
attached internally to each record in DataStream API. When you call
`y.rowtime` during a stream to table conversion, the runtime will
extra
o/pravega/connectors/flink/FlinkPravegaTableITCase.java#L310
Best Regards,
Brian
*From:* Jark Wu
*Sent:* Thursday, March 19, 2020 20:25
*To:* Till Rohrmann
*Cc:* Zhou, Brian; Timo Walther; Jingsong Li; user
*Subject:* Re: Need help on timestamp type conversion for Table API on
Pravega Connect
in the 2nd
time window because it is from the first table and 3rd one.
-- Vinod
On Fri, Mar 13, 2020 at 6:42 AM Timo Walther <mailto:twal...@apache.org>> wrote:
Hi Vinod,
I cannot spot any problems in your SQL query.
Some questions for clarification:
1) Whic
Hi Vinod,
I cannot spot any problems in your SQL query.
Some questions for clarification:
1) Which planner are you using?
2) How do you create your watermarks?
3) Did you unit test with only parallelism of 1 or higher?
4) Can you share the output of TableEnvironment.explain() with us?
Shouldn't
o I prevent this
undesirable behaviour? I'm quite happy for my solution to serialize only
what I explicitly tell it to, I don't need exactly once or anything.
Many thanks,
Chris Stevens
Head of Research & Development
+44 7565 034 595
On Wed, 19 Feb 2020 at 12:19, Timo Walther <
orking in hosted Kinesis
Date: Wed, 19 Feb 2020 12:02:16 +
From: Chris Stevens
To: Timo Walther
Hi Timo,
Thanks for your reply. This makes sense to me, how do I treat something
as a POJO instead of a generic serialized BB type? Sorry relatively new
to Java and Flink.
ch will end up in this
metadata loss. We know what fields the Table must be composed of, but we just
won't know which index they live in so Row#getField() isn't what quite what we
need.
// ah
-Original Message-
From: Timo Walther
Sent: Friday, January 17, 2020 11:29 AM
To:
Hi,
would Apache Avro be an option for you? Because this is currently still
the best supported format when it comes to schema upgrades as far as I
know. Maybe Gordon in CC can give your some additional hints.
Regards,
Timo
On 18.02.20 10:38, ApoorvK wrote:
I have some case class which have
Hi Chris,
it seems there are field serialized into state that actually don't
belong there. You should aim to treat Sensor as a POJO instead of a Kryo
generic serialized black-box type.
Furthermore, it seems that field such as
"org.apache.logging.log4j.core.layout.AbstractCsvLayout" should no
Hi Austin,
the StreamingFileSink allows bucketing the output data.
This should help for your use case:
https://ci.apache.org/projects/flink/flink-docs-stable/dev/connectors/streamfile_sink.html#bucket-assignment
Regards,
Timo
On 19.02.20 01:00, Austin Cawley-Edwards wrote:
Following up on th
Hi Theo,
there are lot of performance improvements that Flink could do but they
would further complicate the interfaces and API. They would require deep
knowledge of users about the runtime when it is safe to reuse object and
when not.
The Table/SQL API of Flink uses a lot of these optimizat
Hi Anuj,
another option would be to use the new Hive connectors. Have you looked
into those? They might work on SQL internal data types which is why you
would need to use the Table API then.
Maybe Bowen in CC can help you here.
Regards,
Timo
On 19.02.20 11:14, Rafi Aroch wrote:
Hi Anuj,
I
Hi,
the type system is still under heavy refactoring that touches a lot of
interfaces. Where would you like to use java.sql.Timestamp? UDFs are not
well supported right now. Source and sinks might work for the Blink
planner and java.sql.Timestamp is the only supported conversion class of
old
Congratualations everyone! Great stuff :-)
Regards,
Timo
On 12.02.20 16:05, Leonard Xu wrote:
Great news!
Thanks everyone involved !
Thanks Gary and Yu for being the release manager !
Best,
Leonard Xu
在 2020年2月12日,23:02,Stephan Ewen 写道:
Congrats to us all.
A big piece of work, nicely don
so quickly!
We could do it with insert-only rows. When you say flags in the data do you
mean a field with a name like 'retracts' and then the value of that field is
the id of the event/row we want to retract? How would that be possible with
Flink?
Thanks!
On 2020/02/04 15:27:20, Timo
Hi Stephan,
the use cases you are describing sound like a perfect fit to Flink.
Internally, Flink deals with insertions and deletions that are flowing
through the system and can update chained aggregations and complex queries.
The only bigger limitation at the moment is that we only support s
Hi,
timestamps and watermarks are attached to every stream record in the
runtime. After assignTimestampsAndWatermarks() extracted them, Flink
handles those attributes internally.
For example, it doesn't matter which class you are emitting in a flatMap
function, the runtime will set the times
Hi Morgan,
as far as I know this is not possible mostly because measuring "till the
point when the system catches up to the last message" is very
pipeline/connector dependent. Some sources might need to read from the
very beginning, some just continue from the latest checkpointed offset.
Mea
Hi Hemant,
maybe this thread from last year could also help you:
http://mail-archives.apache.org/mod_mbox/flink-user/201903.mbox/%3c2df93e1c-ae46-47ca-9c62-0d26b2b3d...@gmail.com%3E
Someone also proposes a skeleton of the code there.
Regards,
Timo
On 04.02.20 08:10, hemant singh wrote:
Than
Hi Andreas,
if dataset.getType() returns a RowTypeInfo you can ignore this log
message. The type extractor runs before the ".returns()" but with this
method you override the old type.
Regards,
Timo
On 15.01.20 15:27, Hailu, Andreas wrote:
Dawid, this approach looks promising. I’m able to fl
201 - 300 of 683 matches
Mail list logo