This in general is not a good idea, as the state you query using queryable
state within a job does not provide any consistency guarantees at all.
Would it be possible to have some trigger that emits state of the windows,
and join the states downstream?
In general, that is a better approach for
Hi,
That in general is not a good idea, with the problem you mentioned as well
as the fact that the state you
query within the same job using queryable state does not provide any means
of consistency guarantee.
When it comes to "querying state from another operator", it is a hint that
your use
Hi Joe,
The main effect this should have is more state to be kept until the windows
can be fired (and state purged).
This would of course increase the time it takes to checkpoint the operator.
I'm not sure if there will be significant runtime per-record impact caused
by how windows are
Are you getting an exception from running the Harness?
The Harness should already have the required configurations, such as the
parent first classloading config.
Otherwise, if you would like to add your own configuration, use the
`withConfiguration` method on the `Harness` class.
On Fri, May 22,
Sorry, forgot to cc user@ as well in the last reply.
On Fri, May 22, 2020 at 12:01 PM Tzu-Li (Gordon) Tai
wrote:
> As an extra note, the utilities you will find in `statefun-e2e-tests`,
> such as the `StatefulFunctionsAppsContainers` is not yet intended for users.
> This however was p
Hi,
On Fri, May 22, 2020 at 7:19 AM Boris Lublinsky <
boris.lublin...@lightbend.com> wrote:
> Also, where do I put flint-conf.yaml in Idea to add additional required
> config parameter:
>
> classloader.parent-first-patterns.additional:
>
Hi Xiaolong,
You are right, the way the Kinesis connector is implemented / the way the
AWS APIs are used, does not allow it to consume Kinesis streams with
enhanced fan-out enabled consumers [1].
Could you open a JIRA ticket for this?
As far as I can tell, this could be a valuable contribution to
In that case, the most possible cause would be
https://issues.apache.org/jira/browse/FLINK-16313, which is included in
Flink 1.10.1 (to be released)
The release candidates for Flink 1.10.1 is currently ongoing, would it be
possible for you to try that out and see if the error still occurs?
On
Hi,
Correct me if I'm wrong, but from the discussion so far it seems like what
Wouter is looking for is an HTTP-based ingress / egress.
We have been thinking about this in the past. The specifics of the
implementation is still to be discussed, but to be able to ensure
exactly-once processing
As others have mentioned already, it is true that method calls on operators
(e.g. processing events and snapshotting state) will not concurrently
happen.
As for your findings in reading through the documentation, that might be a
hint that we could add a bit more explanation mentioning this.
Could
Hi,
The last time I saw this error, was that there was a mismatch in the used
flink-state-processor-api version and other core Flink dependencies.
Could you confirm that?
Also, are you seeing this assertion error consistently, or only
occasionally?
cc'ing Seth, maybe he has other clues on the
Hi Omid,
There currently aren't any benchmarks that I know of for Stateful Functions.
However, Stateful Functions applications run on top of Apache Flink and
therefore share the same network stack / runtime. So, if throughput and
latency is your only concern, you should be able carry over any
Hi,
Could you briefly describe what you are trying to achieve?
By definition, a GlobalWindow includes all data - the ending timestamp for
these windows are therefore Long.MAX_VALUE. An event time trigger wouldn't
make sense here, since that trigger would never fire (watermark can not pass
the
Hi,
As you mentioned, Gelly Graph's are backed by Flink DataSets, and therefore
work primarily on static graphs. I don't think it'll be possible to
implement incremental algorithms described in your SO question.
Have you tried looking at Stateful Functions, a recent new API added to
Flink?
It
Hi!
It isn't possible to use Flink CEP within Stateful Functions.
That could be an interesting primitive, to add CEP-based function
constructs.
Could your briefly describe what you are trying to achieve?
On the other hand, there are plans to integrate Stateful Functions more
closely with the
Hi Oytun!
You can see here an example of how to package a StateFun application image
that contains multiple modules:
https://ci.apache.org/projects/flink/flink-statefun-docs-stable/deployment-and-operations/packaging.html#images
Essentially, for each module you want to include in your
The Apache Flink community is very happy to announce the release of Apache
Flink Stateful Functions 2.0.0.
Stateful Functions is an API that simplifies building distributed stateful
applications.
It's based on functions with persistent state that can interact dynamically
with strong consistency
assOf[TestDataNested]"
> in my current application, basically replace the serialise it throws the
> "new state serialiser is not compaitable.
>
> What can I do here, would be great help thanks in advance
>
> On Fri, Mar 27, 2020 at 1:19 PM Tzu-Li (Gordon) Tai
> wrote
.com/jobs?utm_source=signature_medium=email>
>
> On Mon, Mar 30, 2020 at 1:04 AM Tzu-Li (Gordon) Tai
> wrote:
>
>> It seems like Seth's reply didn't make it to the mailing lists somehow.
>> Forwarding his reply below:
>>
>> -- Forward
It seems like Seth's reply didn't make it to the mailing lists somehow.
Forwarding his reply below:
-- Forwarded message -
From: Seth Wiesman
Date: Thu, Mar 26, 2020 at 5:16 AM
Subject: Re: Lack of KeyedBroadcastStateBootstrapFunction
To: Dawid Wysakowicz
Cc: , Tzu-Li (Gordon
ding.state.CustomAvroSerializer$class.deserialize(CustomAvroSerializer.scala:42)
> at
> nl.mrooding.state.ProductSerializer.deserialize(ProductSerializer.scala:9)
> at
> nl.mrooding.state.ProductSerializer.deserialize(ProductSerializer.scala:9)
> at
> org.apache.flink.contrib.streaming.state
Hi Apoorv,
Flink currently does not natively support schema evolution for state types
using Scala case classes [1].
So, as Roman has pointed out, there are 2 possible ways for you to do that:
- Implementing a custom serializer that support schema evolution for your
specific Scala case classes,
Hi,
The exception stack you posted simply means that the next operator in the
chain failed to process the output watermark.
There should be another exception, which would explain why some operator
was closed / failed and eventually leading to the above exception.
That would provide more insight
Hi,
I believe that the title of this email thread was a typo, and should be
"Apache Flink - Question about checkpointing and re-run a job."
I assume this because the contents of the previous conversations seem to be
purely about Flink.
Otherwise, as far as I know, there doesn't seem to be any
Hi,
As David already explained, they are similar in that you may output zero to
multiple records for both process and flatMap functions.
However, ProcessFunctions also expose to the user much more powerful
functionality, such as registering timers, outputting to side outputs, etc.
Cheers,
Hi John,
Have you considered letting the BroadcastProcessFunction output events that
indicate extra external HTTP requests needs to be performed, and have them
consumed by a downstream async IO operator to complete the HTTP request?
That could work depending on what exactly you need to do in your
Hi Jacob,
Apart from what Klou already mentioned, one slightly possible reason:
If you are using the FsStateBackend, it is also possible that your state is
small enough to be considered to be stored inline within the metadata file.
That is governed by the "state.backend.fs.memory-threshold"
Hi,
Flink currently performs a 128-bit murmur hash on the user-provided uids to
generate the final node hashes in the stream graph. Specifically, this
library is being used [1] as the hash function.
If what you are looking for is for Flink to use exactly the provided hash,
you can use
Hi,
What that LOG means (i.e. "must be processed as a Generic Type") is that
Flink will have to fallback to using Kryo for the serialization for that
type.
You should be concerned about that if:
1) That type is being used for some persisted state in snapshots. That would
be the case if you've
Hi Kaymak,
To answer your last question:
there will be no data loss in that scenario you described, but there could
be duplicate processed records.
With checkpointing enabled, the Flink Kafka consumer does not commit
offsets back to Kafka until offsets in Flink checkpoints have been
persisted.
Hi,
The connectors that are listed in the AWS documentation page that you
referenced are not provided by AWS. They are bundled connectors shipped by
the Apache Flink community as part of official Flink releases, and are
discoverable as artifacts from the Maven central repository. See the
Hi,
First of all, state is only managed by Flink (and therefore Flink's state
backends) if the state is registered by the user.
You can take a look at the documents here [1] on details on how to register
state.
A state has to be registered for it to be persisted in checkpoints /
savepoints, and
Hi Laurent,
You can take a look at Flink's MiniClusterResource JUnit test rule, and its
usages in the codebase for that.
The rule launches a Flink MiniCluster within the same JVM, and submission to
the mini cluster resembles how it would be submitting to an actual Flink
cluster, so you would
Hi Theo,
This is indeed a tricky feature to test!
On Thu, Feb 20, 2020 at 8:59 PM Theo Diefenthal <
theo.diefent...@scoop-software.de> wrote:
> Hi,
>
> We have a pipeline which internally uses Java POJOs and also needs to keep
> some events entirely in state for some time.
>
> From time to
is that type
information extraction should be converged for the Java / Scala DataStream
APIs.
Cheers,
Gordon
On Wed, Feb 19, 2020 at 10:20 AM Tzu-Li (Gordon) Tai
wrote:
> Hi,
>
> Just to clarify -
> I quickly went through the README of the project, and saw this:
> "This
Hi,
Just to clarify -
I quickly went through the README of the project, and saw this:
"This error is seen after trying to read from a savepoint that was created
using the same case class as a key."
So, if I understood correctly, you were attempting to use the State
Processor API to access a
Hi,
There are no upper limits on the number of Kafka consumers per job.
For each one of your FlinkKafkaConsumers, are you using the same group.id?
That could maybe explain why you are experiencing higher commit times as
you are adding more FlinkKafkaConsumers, as AFAIK on the broker side, the
any prior snapshotted state (i.e. offsets) and respect the startup
configuration.
Let me know if this works for you!
Cheers,
Gordon
On Thu, Jan 23, 2020 at 9:12 PM Tzu-Li (Gordon) Tai
wrote:
> Hi Somya,
>
> I'll have to take a closer look at the JIRA history to refresh my memory
> on
Hi,
If your filter and flatMap operators are chained, then the performance
difference should not be noticeable.
If a shuffle (i.e. a keyBy operation) occurs after the filter and before
the flatMap, then applying the filter first will be more efficient.
Cheers,
Gordon
On Thu, Jan 30, 2020 at
Hi Ran,
On Thu, Jan 30, 2020 at 9:39 AM Ran Zhang wrote:
> Hi all,
>
> We have a Flink app that uses a KeyedProcessFunction, and in the function
> it requires a ValueState(of TreeSet) and the processElement method needs to
> access and update it. We tried to use RocksDB as our stateBackend but
Hi Somya,
I'll have to take a closer look at the JIRA history to refresh my memory on
potential past changes that caused this.
My first suspection is this:
It is expected that the Kafka consumer will *ignore* the configured startup
position if the job was restored from a savepoint.
It will
Hi Izual,
Thanks for reporting this! I'm also forwarding this to the user mailing
list, as that is the more suitable place for this question.
I think the usability of the State Processor API in Scala is indeed
something that hasn’t been looked at closely yet.
On Tue, Jan 21, 2020 at 8:12 AM
Hi Vasily,
State names need to be unique within operators only.
Cheers,
Gordon
On Mon, Jan 20, 2020 at 10:58 AM Vasily Melnik <
vasily.mel...@glowbyteconsulting.com> wrote:
> Hi all,
>
> I'm a bit confused with state name uniqueness.
> Should it be unique within operator only, or within entire
Hi Bekir,
Before diving deeper, just to rule out the obvious:
Have you changed anything with the element type of the input stream to the
async wait operator?
This wasn't apparent from the information so far, so I want to quickly
clear that out of the way first.
Cheers,
Gordon
On Wed, Oct 30,
The Apache Flink community is very happy to announce the release of Apache
Flink 1.9.0, which is the latest major release.
Apache Flink® is an open-source stream processing framework for
distributed, high-performing, always-available, and accurate data streaming
applications.
The release is
-- Forwarded message -
From: Tzu-Li (Gordon) Tai
Date: Tue, Aug 20, 2019 at 1:16 AM
Subject: [VOTE] Apache Flink 1.9.0, release candidate #3
To: dev
Hi all,
Release candidate #3 for Apache Flink 1.9.0 is now ready for your review.
Please review and vote on release candidate #3 for version
-- Forwarded message -
From: Tzu-Li (Gordon) Tai
Date: Fri, Aug 9, 2019 at 6:17 PM
Subject: [VOTE] Apache Flink Release 1.9.0, release candidate #2
To: dev
Hi all,
Release candidate #2 for Apache Flink 1.9.0 is now ready for your review.
This is the first voting candidate for 1.9.0
; operator state through CheckpointedFunction interface and this sink isn’t
> used in all our jobs.
>
> Med venlig hilsen / Best regards
> Lasse Nedergaard
>
>
> Den 3. jun. 2019 kl. 12.50 skrev Tzu-Li (Gordon) Tai >:
>
> Hi Lasse,
>
> This is indeed a bit odd. I
I quickly checked the implementation of duplicate() for both the
KryoSerializer and StreamElementSerializer (which are the only serializers
involved here).
They seem to be correct; especially for the KryoSerializer, since
FLINK-8836 [1] we now always perform a deep copy of the KryoSerializer when
Hi Andrea,
Is there a specific reason you want to use a custom TypeInformation /
TypeSerializer for your type?
>From the description in the original post, this part wasn't clear to me.
If the only reason is because it is generally suggested to avoid generic
type serialization via Kryo, both for
pache.org/news/2019/07/02/release-1.8.1.html
>>>
>>> The full release notes are available in Jira:
>>>
>>> https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12315522=12345164
>>>
>>> We would like to thank all contributors of the Apache Flink community
>>> who made this release possible!
>>>
>>> Great thanks to @Tzu-Li (Gordon) Tai 's offline
>>> kind help!
>>>
>>> Regards,
>>> Jincheng
>>>
>>
>
Reportmessage extends ReportmessageBase and the state operator use
> ReportmessageBase.
> So we need to register all the class’s that extends a class used in state.
> Don’t know why this is needed in 1.8
>
> Med venlig hilsen / Best regards
> Lasse Nedergaard
>
>
> Den 28. maj 201
FYI: Seth starting a FLIP for adding a savepoint connector that addresses
this -
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/Discuss-FLIP-43-Savepoint-Connector-td29233.html
Please join the discussion there if you are interested!
On Thu, Mar 28, 2019 at 5:23 PM Tzu-Li (Gordon
Hi Lasse,
Did you move the class to a different namespace / package or changed to be
a nested class, across the Flink versions?
That would be the only cause I could reason about at the moment.
If possible, could you also have a very minimal snippet / instructions on
how I can maybe reproduce
Hi Burgess,
Would you be able to provide a minimal project that can reproduce your
error?
That would help a lot with figuring out the issue.
If you prefer to share that only privately, please feel free to send me a
private email with the details.
Another thing you can do is set logging level to
Hi Steven,
I assume you are referring to the problem that we don't publish the Kinesis
connector artifacts to Maven, due to the licensing issue with KCL?
I didn't manage to find any JIRAs that were addressing this issue directly,
but the most related one would be this:
Hi,
Just to add to what Piotr already mentioned:
The community is working on adding support for this directly in Flink.
You can follow the efforts here:
https://issues.apache.org/jira/browse/FLINK-12047.
Cheers,
Gordon
On Tue, May 14, 2019 at 11:39 AM Piotr Nowojski wrote:
> Hi,
>
>
Hi,
Aljoscha opened a JIRA just recently for this issue:
https://issues.apache.org/jira/browse/FLINK-12501.
Do you know if this is a regression from previous Flink versions?
I'm asking just to double check, since from my understanding of the issue,
the problem should have already existed before.
Hi Marc!
I know we talked offline about the issues mentioned in this topic already,
but I'm just relaying the result of the discussions here to make it
searchable by others bumping into the same issues.
On Thu, Mar 21, 2019 at 4:27 PM Marc Rooding wrote:
> Hi
>
> I’ve been trying to get state
hema evolution still be triggered? Or does it actually go down to the
> avro schema rather than just the class serialVersionUID?
>
>
>
>
>
>
> On Mon, Mar 18, 2019 at 1:10 AM Tzu-Li (Gordon) Tai
> wrote:
>
>> Hi Cliff,
>>
>> Thanks for bringing th
@Ufuk
Yes, creating a JIRA now already to track this makes sense.
I've proceeded to open one:
https://issues.apache.org/jira/browse/FLINK-12047
Let's move any further discussions there.
Cheers,
Gordon
On Thu, Mar 28, 2019 at 5:01 PM Ufuk Celebi wrote:
> I think such a tool would be really
Hi,
Do you have the full error message of the failure?
A wild guess to begin with: have you made sure that there are sufficient
permissions to create the directory?
Best,
Gordon
On Tue, Mar 26, 2019 at 5:46 PM Paul Lam wrote:
> Hi,
>
> I have a job (with Flink 1.6.4) which uses rocksdb
Hi!
Regarding the support for savepoint reading / writing / processing directly
in core Flink, we've been thinking about that lately and might push a bit
for adding the functionality to Flink in the next release.
For example, beside Bravo, Seth (CC'ed) also had implemented something [1]
for this.
Hi Cliff,
Thanks for bringing this up!
AFAIK, this wouldn't be a long-term blocker. I've just opened a JIRA to
track this [1].
As explained in the JIRA ticket, the main reason this is disallowed in the
initial support for state schema evolution was due to how migration was
implemented in the
Hi,
Currently Flink uses Kryo as the default serializer for data types that
Flink's type serialization stack doesn't support [1].
This also includes serializers being used for managed state registered by
users.
Because of this, at the moment it's not easy to upgrade the Kryo version,
since it is
Hi Pavel,
As you already discovered, this problem occurs still because in 1.7.x, the
KryoSerializer is still using the deprecated TypeSerializerConfigSnapshot
as its snapshot, which relies on the serializer being Java-serialized into
savepoints as state metadata.
In 1.8.0, all Flink's built-in
ngOutput.collect(OperatorChain.java:554)
> at
> org.apache.flink.streaming.runtime.tasks.OperatorChain$CopyingChainingOutput.collect(OperatorChain.java:534)
> at
> org.apache.flink.streaming.api.operators.AbstractStreamOperator$CountingOutput.collect(AbstractStreamOperator.java:718)
>
>
&
Hi,
The Apache Flink community is very happy to announce the release of
Apache Flink 1.7.2, which is the second bugfix release for the Apache
Flink 1.7 series.
Apache Flink® is an open-source stream processing framework for
distributed, high-performing, always-available, and accurate data
Hi,
@Averell I renamed the
`ElasticsearchFailureHandlerIndexer` to be `BufferingNoOpRequestIndexer`,
which explains why you can't find it.
The voting thread for RC#1 of 1.7.2 can be found at [1].
The actual commits which fixes the problem are d9c45af to 2f52227.
Cheers,
Gordon
[1]
Hi,
除了 Apache Flink 官方文件以外 [1],我個人也建議可以看看 Ververica 這一系列的 Flink training 題材:
https://training.ververica.com/
除此之外,學習過程中有遇到任何問題也歡迎可以直接發信件跟我們詢問。
- Gordon
[1] https://flink.apache.org/
On Thu, Feb 14, 2019 at 11:44 AM shen lei wrote:
> 有木有好的经验或者方法分享一下,感谢。最近学的,感觉还是不系统。
Thanks for testing it out.
Will be great to get your feedback on whether the release candidate for
1.7.2 fixes this for you.
On Wed, Feb 13, 2019 at 7:38 PM Averell wrote:
> Thank you Gordon.
>
> That's my exact problem. Will try the fix in 1.7.2 now.
>
> Thanks and regards,
> Averell
>
>
>
>
Hi,
Which Flink version are you using?
Cheers,
Gordon
--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
Hi,
I think this is unexpected. The generated transactional ids should not be
clashing.
Looking at the FlinkKafkaProducer code, it seems like the generation is
only a function of the subtask id of the FlinkKafkaProducer, which could be
the same across 2 different Kafka sources.
I'm not
Hi,
I'm forwarding this question to Gary (CC'ed), who most likely would have an
answer for your question here.
Cheers,
Gordon
On Wed, Feb 13, 2019 at 8:33 AM Jins George wrote:
> Hello community,
>
> I am trying to upgrade a Flink Yarn session cluster running BEAM
> pipelines from version
Hi Alexey,
I don't think using the @TypeInfo annotation is doable at the moment.
Is this class being used only for input / output types of functions /
operators?
Or are you using it as a state type?
For the former, I think you can explicitly set the TypeInformation by
calling setTypeInfo on
Hi,
I would suggest to avoid Kryo for state serialization, especially if this
job is meant for production usage.
It might get in the way in the future when you might decide to upgrade your
value state schema.
To do that, when declaring the descriptor for your MapState, provide a
specific
Hi,
If Firehouse already supports sinking records from a Kinesis stream to an
S3 bucket, then yes, Chesnay's suggestion would work.
You route each record to a specific Kinesis stream depending on some value
in the record using the KinesisSerializationSchema, and sink each Kinesis
stream to their
Hi Averell,
This seems to be the bug that you encountered:
https://issues.apache.org/jira/browse/FLINK-11046.
Cheers,
Gordon
On Sat, Feb 9, 2019 at 3:27 PM Averell wrote:
> Hello,
>
> I am trying to follow this Flink guide [1] to handle errors in
> ElasticSearchSink by re-adding the failed
Hi!
Thanks for reporting this.
This looks like a bug that we fixed in Flink 1.7.1 [1].
Would you be able to try with 1.7.1 and see if the issue is still happening
for you?
Cheers,
Gordon
[1] https://issues.apache.org/jira/browse/FLINK-11094
On Tue, Jan 29, 2019, 6:29 PM Averell I tried to
Hi,
Yes, Dawid is correct.
The "group.id" setting in Flink's Kafka Consumer is only used for group
offset fetching and committing offsets back to Kafka (only for exposure
purposes, not used for processing guarantees).
The Flink Kafka Consumer uses static partition assignment on the
KafkaConsumer
Hi!
Yes, Flink's watermark timestamps are in milliseconds, which means that
time-based operators such as time window operators will be fired at a
per-millisecond granularity.
Whether or not this introduces "latency" in the pipeline depends on the
granularity of your time window operations; if you
Hi Sohi!
On Wed, Jan 23, 2019 at 9:01 PM sohimankotia wrote:
> Hi,
>
> Let's say I have flink Kafka consumer read from 3 topics , [ T-1 ,T-2,T-3
> ]
> .
>
> - T1 and T2 are having partitions equal to 100
> - T3 is having partitions equal to 60
> - Flink Task (parallelism is 50)
>
There isn't
Thanks Peter! Yes, it would also be great if you try the patch in
https://github.com/apache/flink/pull/7580 out and see if that works for you.
On Mon, Jan 28, 2019 at 7:47 PM pwestermann
wrote:
> Hi Gordon,
>
> We should be able to wait for 1.7.2 but I will also test the workaround and
> post
Hi,
Thanks for all the information and reporting this.
We've identified this to be an actual issue:
https://issues.apache.org/jira/browse/FLINK-11436.
There's also a PR opened to fix this, and is currently under review:
https://github.com/apache/flink/pull/7580.
I'll make sure that this is fixed
Hi!
We've double checked the code, and the only plausible cause of this is that
you may be using flink-avro 1.6.x with Flink 1.7.x.
Could you double check that all Flink dependencies, including flink-avro,
are 1.7.1?
You can verify this by doing `mvn dependency:tree` on your job, and check
that
Thanks for the logs.
Is the job restore actually failing? If yes, there should be an exception
for the exact cause of the failure.
Otherwise, the AvroSerializer warnings in the taskmanager logs is actually
expected behaviour when restoring from savepoint versions before 1.7.x, and
shouldn't
Hi,
Thanks for reporting this.
Could you provide more details (error message, exception stack trace) that
you are getting?
This is unexpected, as the changes to flink-avro serializers in 1.7.x
should be backwards compatible.
More details on how the restore failed will be helpful here.
Cheers,
Hi,
Have you looked at [1]?
You can annotate your type and provide a type info factory. The factory
would be used to create the TypeInformation for that type, and in turn
create the serializer used for that type.
[1]
Hi,
The Apache Flink community is very happy to announce the release of
Apache Flink 1.6.3, which is the third bugfix release for the Apache
Flink 1.6 series.
Apache Flink® is an open-source stream processing framework for
distributed, high-performing, always-available, and accurate data
> that falls back to Kryo will cause an error.
>
> Would love to see more updated slides if you don't mind.
>
> Thanks for taking the time,
> Padarn
>
>
> On Fri, Dec 21, 2018 at 10:04 PM Tzu-Li (Gordon) Tai
> wrote:
>
>> For the documents I would recommend re
For the documents I would recommend reading through:
https://ci.apache.org/projects/flink/flink-docs-release-1.7/dev/stream/state/schema_evolution.html
On Fri, Dec 21, 2018, 9:55 PM Tzu-Li (Gordon) Tai Hi,
>
> Yes, if Flink does not recognize your registered state type, it will by
> de
Hi,
Yes, if Flink does not recognize your registered state type, it will by
default use Kryo for the serialization.
And generally speaking, Kryo does not have good support for evolvable
schemas compared to other serialization frameworks such as Avro or Protobuf.
The reason why Flink defaults to
the connections to elastic cluster reached to:
netstat -aln | grep 9200 | wc -l
2333
On Thu, Dec 13, 2018 at 4:12 PM Tzu-Li (Gordon) Tai wrote:
Hi,
Besides the information that Chesnay requested, could you also provide a stack
trace of the exception that caused the job to terminate in the first place
issues with LATEST option.
TIA,
Vijay
On Thu, Dec 13, 2018 at 2:59 AM Tzu-Li (Gordon) Tai wrote:
Hi!
Thanks for reporting this.
This looks like an overlooked corner case that the Kinesis connector doesn’t
handle properly.
First, let me clarify the case and how it can be reproduced. Please let
Hi!
Thanks for reporting this.
This looks like an overlooked corner case that the Kinesis connector doesn’t
handle properly.
First, let me clarify the case and how it can be reproduced. Please let me know
if the following is correct:
1. You started a Kinesis connector source, with
Hi,
Besides the information that Chesnay requested, could you also provide a stack
trace of the exception that caused the job to terminate in the first place?
The Elasticsearch sink does indeed close the internally used Elasticsearch
client, which should in turn properly release all resources
roxy-chain...com
> kinesisConsumerConfig.setProperty(AWSUtil.AWS_CLIENT_CONFIG_PREFIX +
> "proxyHost", proxyHost);//<== mo http:// in proxyHost name
>
> TIA,
> Vijay
>
>
> On Wed, Nov 14, 2018 at 12:50 AM Tzu-Li (Gordon) Tai
> wrote:
>
>> Hi Vijay,
>>
>> I’m prett
Hi Henry,
Flushing of buffered data in sinks should occur on two occasions - 1) when some
buffer size limit is reached or a fixed-flush interval is fired, and 2) on
checkpoints.
Flushing any pending data before completing a checkpoint ensures the sink has
at-least-once guarantees, so that
Hi,
Have you taken a look yet at this [1]? That is an example of writing a stream
to HBase.
Cheers,
Gordon
[1]
https://github.com/apache/flink/blob/master/flink-connectors/flink-hbase/src/test/java/org/apache/flink/addons/hbase/example/HBaseWriteStreamExample.java
On 11 November 2018 at
Hi,
AFAIK, I don’t think there has been other discussions on this other than the
original document on secured data access for Flink [1].
Unfortunately, I’m also not knowledgeable enough to comment on how feasible it
would be to support MD5-Digest for authentication.
Maybe Eron (cc’ed) can
101 - 200 of 546 matches
Mail list logo