his is indeed true. We had some plans for how to resolve this but
> they never materialised because of the focus on Stream Processing. We might
> unite the two in the future and then you will get fault-tolerant
> batch/stream processing in the same API.
>
>
>
> Best,
>
> Aljoscha
>
Hi, all:
I'm learning flink's doc and curious about the fault tolerance of batch
process jobs. It seems that when one of task execution fails, the whole job
will be restarted, is it true? If so, isn't it impractical to deploy large
flink batch jobs?
--
Liu, Renjie
Software Engineer, MVAD
Essentially you are right, but the snapshot commit process is asynchronous.
That's what you have to pay for exactly once semantics.
Li Wang 于2016年11月1日周二 下午3:05写道:
> Hi all,
>
> I have a question regarding to the state checkpoint mechanism in Flink. I
> find the statement
Sorry the incorrect reply, please ignore this.
On Tue, Nov 1, 2016 at 8:47 PM Renjie Liu <liurenjie2...@gmail.com> wrote:
> Essentially you are right, but the snapshot commit process is
> asynchronous. That's what you have to pay for exactly once semantics.
>
> Li Wang <w
Hi, Till:
By operator with multiple inputs, do you mean inputs from multiple
subtasks?
On Tue, Nov 1, 2016 at 8:56 PM Till Rohrmann wrote:
> Hi Li,
>
> the statement refers to operators with multiple inputs (two in this case).
> With the current implementation you will
Thanks for the reply.
On Wed, Nov 2, 2016 at 5:19 PM Till Rohrmann <trohrm...@apache.org> wrote:
> Yes you're right. Whenever you have multiple input channels which could
> also be the case if you do a repartitioning between two mappers.
>
> On Tue, Nov 1, 2016 at 11
Hi, all:
It seems that flink's checkpoint mechanism saves state per partition.
However, if I want to change configuration of the job and resubmit the job,
I can not recover from last checkpoint, right? Say I change the parallelism
of kafka consumer and partitions assigned to each subtask will be
ferent parallelism.
>
> 2016-11-04 11:24 GMT+01:00 Renjie Liu <liurenjie2...@gmail.com>:
>
> Hi, all:
> It seems that flink's checkpoint mechanism saves state per partition.
> However, if I want to change configuration of the job and resubmit the job,
> I can not reco
t;
> With multiple inputs I referred to a coFlatMap operator or a join operator
> which have both two inputs.
>
> Cheers,
> Till
>
> On Tue, Nov 1, 2016 at 3:29 PM, Renjie Liu <liurenjie2...@gmail.com>
> wrote:
>
> Hi, Till:
> By operator with multiple inputs, do you me
Great!
On Wed, Dec 14, 2016 at 1:30 AM Matthias J. Sax wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA512
>
> I think it's worth to announce this via news list. :)
>
> On 12/13/16 7:32 AM, Robert Metzger wrote:
> > The commun...@flink.apache.org
I don’t think this should be a issue
> for the 09 consumer, since we’re solely relying on the Kafka client’s own
> implementation to do the auto offset committing.
>
> Cheers,
> Gordon
>
>
> On January 9, 2017 at 7:55:33 PM, Timo Walther (twal...@apache.org) wrote:
>
>
Hi, all:
I'm using flink 1.1.3 and kafka consumer 09. I read its code and it says
that the kafka consumer will turn on auto offset commit if checkpoint is
not enabled. I've turned off checkpoint and it seems that kafka client is
not committing to offsets to kafka? The offset is important for
gt; On Thu, Mar 23, 2017 at 11:38 AM, Renjie Liu <liurenjie2...@gmail.com>
> wrote:
>
> Permanent. I've waited for several minutes and the task manager is still
> lost.
>
> On Thu, Mar 23, 2017 at 6:34 PM Ufuk Celebi <u...@apache.org> wrote:
>
> When it happ
We are running both systems. Old existing systems are still running on
storm and new jobs are all running on flink.
On Mon, Aug 21, 2017 at 3:06 AM kaniska Mandal
wrote:
> Storm supports 'exactly once' -
> http://storm.apache.org/releases/current/Trident-tutorial.html
Hi, does this include the FLIP6?
On Tue, May 9, 2017 at 2:29 AM Stephan Ewen wrote:
> Did a quick test: Simply adding the
> "org.apache.maven.plugins.shade.resource.ApacheNoticeResourceTransformer"
> helps with NOTICE files,
> but does not add the required BSD licence copies.
ue, May 9, 2017 at 10:02 AM, Renjie Liu <liurenjie2...@gmail.com>
> wrote:
>
>> Stephan:
>> This means that we can use FLIP 6 for basic functionality now? Any
>> document for setting up that? Will FLIP 6 be finished in release 1.3?
>>
>> On Tu
Stephan:
This means that we can use FLIP 6 for basic functionality now? Any
document for setting up that? Will FLIP 6 be finished in release 1.3?
On Tue, May 9, 2017 at 3:43 PM Stephan Ewen <se...@apache.org> wrote:
> @Renjie Liu: A good part of the FLIP-6 code is in there, bu
@Jason I think there's a mistake in your explanation since each task in the
task manager has its own copy of an operator instance, so the tuple may not
be shared. State is a great solution but I think that's not the root cause.
@Tarek What's the parallelism of your data stream? I think the reason
Jason's solution is right, l'm just clarifying the mistake in the
explanation.
Tarek khal 于2017年5月19日周五 下午7:11写道:
> Hello Renjie,
>
> Yes, the parallelism is 1. what should i do pls ?
>
> Regards,
>
>
>
> --
> View this message in context:
>
Even if you increase the operator parallelism, you can still use the state
operation.
On Fri, May 19, 2017 at 7:47 PM Tarek khal
wrote:
> If I increase the parallelism operator, I risk losing shared state solution
> or it has nothing to do.
> And if it's going to
You need to check that whether you disk is full.
On Thu, Jun 14, 2018 at 2:15 PM Lasse Nedergaard
wrote:
> Hi.
>
> We sometimes see job fails with a blob store exception, like the one
> below. Anyone has an idea why we get them, and how to avoid them?.
> In this case the job has run without any
Could you please paste your sql here?
On Thu, Jun 14, 2018 at 2:54 PM 徐林彬 wrote:
> Hi all
> I'm trying to prove flink sql stream non-windowed inner join with flink
> 1.5.0, but it failed.
> Then i tried flink test
>
Hi:
We've deployed flink 1.5.0 and tested the new cluster manager, it's really
great for flink to be elastic. However we've also found some problems that
blocks us from deploying it to production environment.
1. Task manager isolation. Currently flink allows different jobs to execute
on same
It seems that it's related with your development environment settings.
On Sun, Jul 29, 2018 at 4:39 PM Mich Talebzadeh
wrote:
> Hi,
>
> I have written a simple test program as below
>
> import java.util.Properties
> import java.util.Arrays
> import
Hi,
Another way to ensure order is by adding a logical version number for each
message so that earlier version will not override later version. Timestamp
depends on your ntp server works correctly.
On Sun, Jul 29, 2018 at 3:52 PM Niels Basjes wrote:
> Hi,
>
> The basic thing is that you will
ly Docker container is the one that runs Job Manager.
>
>
>
> What am I missing?
>
>
>
> Thanks,
>
> Alex
>
>
>
> *From:* Renjie Liu [mailto:liurenjie2...@gmail.com]
> *Sent:* Friday, July 20, 2018 8:56 PM
> *To:* Till Rohrmann
> *Cc:* NEKRASSOV, ALEXE
here is an even older
> ticket describing the same problem:
>
> https://issues.apache.org/jira/browse/FLINK-7470
>
> One of them should be closed.
>
> Best,
> Gary
>
>
> On Thu, Jul 19, 2018 at 10:45 AM, Renjie Liu
> wrote:
>
>> Attached is job ma
Hi, Mich:
You can add write a sink function for that.
On Mon, Jul 30, 2018 at 2:58 PM Mich Talebzadeh
wrote:
>
> Hi,
>
> I have a Kafka topic that transmits 100 security prices ever 2 seconds.
>
> In Spark streaming I go through the RDD and walk through rows one by one
> and check prices
> In
Hi, Alexey:
It depends on the state backend you use. If you use heap memory backend,
then you don't need to do put again.
However, if you use rocksdb state backend, then you need to do the put
again so that it will be saved by the checkpoint.
On Tue, Aug 14, 2018 at 4:58 AM Alexey Trenikhun
Hi, yuvraj:
Do you mean querying hive with sql? Or anything else?
On Tue, Aug 14, 2018 at 3:52 AM yuvraj singh <19yuvrajsing...@gmail.com>
wrote:
> I want to know ,if FLink have support for hive .
>
> Thanks
> Yubraj Singh
>
--
Liu, Renjie
Software Engineer, MVAD
Hi, Stephan:
Can we put project wiki in some place? I think it's a great place to find
flip lists
On Mon, Aug 27, 2018 at 9:56 AM vino yang wrote:
> +1
> The reason is the same as Hequn, because we have given a link to
> SlideShare under the "Flink Forward" section.
>
> Thanks, vino.
>
> Hequn
Hi, Alexei:
What you paste is expected behavior. Jobmanager, two task managers each
should run in a docker instance.
13276 is should be the process of job manager, and it's the same
process as 789.
They have different processes id because in show them in different
namesapces(that's a concept in
Hi, all:
I'm testing flink 1.5.0 and find that flink mesos resource manager unable
to connect to mesos after restart. Have you seen this happenen?
--
Liu, Renjie
Software Engineer, MVAD
are able to re-produce this reliably, can you post the jobmanager
> logs?
>
> Best,
> Gary
> On Wed, Jul 18, 2018 at 10:33 AM, Renjie Liu
> wrote:
>
>> Hi, all:
>>
>> I'm testing flink 1.5.0 and find that flink mesos resource manager unable
>> to conne
You can use rest api here
https://ci.apache.org/projects/flink/flink-docs-release-1.5/monitoring/rest_api.html
On Wed, Jul 25, 2018 at 5:18 PM 从六品州同 <26304...@qq.com> wrote:
> dear all:
> Is there a notification mechanism for Flink? When job's status changes,
> such as restart, failure, notify
ote:
>
> Hi Renjie,
>
> 1. For what I could grasp from Kafka docs, you can subscribe and still use
> poll() to capture a specific offset. But I just read the starting point of
> it and didn't go deep into it.
>
> 2. Currently, Flink 1.4.2, Kafka 0.10.1 and the FlinkKafkaCo
Hi:
1. I'm wondering in which case you need this LAZY_FROM_SROUCE schedule mode?
2. It seems that this can't be changed for streaming jobs, and it's worth a
discussion for an api that changes the schedule mode
On Fri, Sep 7, 2018 at 6:25 PM 陈梓立 wrote:
> Sorry to attach this message here, but if
Hi, kant:
Full support for sql is not easy tuo support. A simple KV storage is not
enough to build a RDBMS.
On Sun, Mar 4, 2018 at 1:42 AM kant kodali wrote:
> Hi Fabian,
>
> Does it make sense to have it on the roadmap? that way external
> applications can do ad-hoc queries
Hi, Chris:
Where are these words from? Are they from flink source code?
On Wed, Oct 10, 2018 at 10:18 PM Chris Miller wrote:
> Hi,
>
> what's the difference between partitions and subpartitions?
>
>
>
> Thanks.
>
>
>
> CM
>
>
--
Renjie Liu
Software Engineer, MVAD
Hi, Julio:
1. Flink doesn't use subscribe because it needs to control partition
assignment itself, which is important for implementing exactly once.
2. Can you share the versions you are using, including kafka, kafka client,
flink? We are also use flink kafka consumer and we can monitor it
Hi, Julio:
This happens frequently? What state backend do you use? The async
checkpoint duration and sync checkpoint duration seems normal compared to
others, it seems that most of the time are spent acking the checkpoint.
On Sun, Sep 16, 2018 at 9:24 AM vino yang wrote:
> Hi Julio,
>
> Yes, it
Hi, Martin:
I think a better solution would be to set the number of cores of each
container equals to that of a physical server if this mesos cluster is
dedicated to your flink cluster.
On Mon, Sep 17, 2018 at 5:28 AM Martin Eden wrote:
> Hi Till,
>
> I was able to use
ing so long?
>
> The checkpoint size of task #8 is twice as big as the second biggest
> checkpoint. But this should not cause an increase in checkpoint time of a
> factor of 8.
>
> Cheers,
> Till
>
> On Mon, Sep 17, 2018 at 5:25 AM Renjie Liu
> wrote:
>
>>
43 matches
Mail list logo