Hi everyone,
I'm getting the following error when trying to restore from a savepoint.
Here below is the output from flink bin, and in the attachment is a TM log.
I didn't have any change in the app before and after savepoint. All Window
operators have been assigned unique ID string.
Could you ple
Hi Averell,
Do you try to scale the job up, meaning do you increase the job
parallelism? Have you increased the job max parallelism by chance? If so
this is not supported. The max parallelism parameter is used to create
key groups that can be further assigned to parallel operators. This
parameter
Hi,
adding to Dawids questions, it would also be very helpful to know which Flink
version was used to create the savepoint, which Flink version was used in the
restore attempt, if the savepoint was moved or modified. Outside of potential
conflicts with those things, I would not expect anything
Hi,
I don’t think that’s exposed on the TaskManager.
Maybe it would simplify things a bit if you implement this as a single
“JobManager” health check, not multiple TaskManagers health check - for example
verify that there are expected number of registered TaskManagers. It might
cover your case
Hi,
Was this happening in older Flink version? Could you post in what circumstances
the job has been moved to a new TM (full job manager logs and task manager logs
would be helpful)? I’m suspecting that those leftover files might have
something to do with local recovery.
Piotrek
> On 9 Oct 2
Hi again,
Glad that you solved your problem :)
Splitting code into smaller functions has its advantages, but more
operators/tasks means more overhead for JobManager/TaskManager to manage them.
Usually that’s not a big issue, but as I said, you were running your cluster on
extremely low memory
Thnx Piotr. I agree, that would work. It's a bit chicken and the egg
problem, since at that point we can't just spin up a task manager, and have
it register itself, we need to have flinkmanager know how many task
managers should be there. Bit more logic, but doable. Thnx for the tip.
Cheers,
Baris
Hi Stefan, Dawid,
I hadn't changed anything in the configuration. Env's parallelism stayed at
64. Some source/sink operators have parallelism of 1 to 8. I'm using Flink
1.7-SNAPSHOT, with the code pulled from master about 5 days back. Savepoint
was saved to either S3 or HDFS (I tried multiple time
Hi Stefan, Dawid,
I hadn't changed anything in the configuration. Env's parallelism stayed at
64. Some source/sink operators have parallelism of 1 to 8. I'm using Flink
1.7-SNAPSHOT, with the code pulled from the master branch about 5 days back.
Savepoint was saved to either S3 or HDFS (I tried mu
You’re welcome :)
> On 10 Oct 2018, at 10:28, Bariša wrote:
>
> Thnx Piotr. I agree, that would work. It's a bit chicken and the egg problem,
> since at that point we can't just spin up a task manager, and have it
> register itself, we need to have flinkmanager know how many task managers
> s
Hi Averell,
In the logs there are some “Split Reader: Custom File Source:”
This is a custom source you implemented?
Also is your keySelector deterministic with proper equals and hashcode methods?
Cheers,
Kostas
> On Oct 10, 2018, at 10:50 AM, Averell wrote:
>
> Hi Stefan, Dawid,
>
> I hadn'
In which method are you calling getRuntimeContext()? This method can
only be used after open() has been called.
On 09.10.2018 17:09, Ahmad Hassan wrote:
Hi,
We want to use MapState inside fold function to keep the map of all
products that we see in 24 hour window to store huge state in rocksd
Yes, it would be good to post your code.
Are you using a FoldFunction in a window (if yes, what window) or as a
running aggregate?
In general, collecting state in a FoldFunction is usually not something
that you should do. Did you consider using an AggregateFunction?
Fabian
Am Mi., 10. Okt. 2018
Hi Xuefu,
Welcome to the Flink community and thanks for starting this discussion!
Better Hive integration would be really great!
Can you go into details of what you are proposing? I can think of a couple
ways to improve Flink in that regard:
* Support for Hive UDFs
* Support for Hive metadata cat
Hi Kostas,
Yes, I modified ContinuousFileMonitoringFunction to add one more
ListState. The error might/should have come from that, but I haven't
been able to find out why.
All of my keyed streams are defined by Scala tuples like: /ikeyBy(r =>
(r.customer_id, r.address))/, and the fields using as
You restore your job with the custom source from a savepoint taken without the
custom source?
> On Oct 10, 2018, at 11:34 AM, Averell wrote:
>
> Hi Kostas,
>
> Yes, I modified ContinuousFileMonitoringFunction to add one more
> ListState. The error might/should have come from that, but I haven
invoke FlinkKafkaProducer011 constructor in scala:
val producer = new
FlinkKafkaProducer011[PVEvent.Entity](appConf.getPvEventTopic, new
PvEventSerializeSchema,
producerProps, Optional.of(FlinkRebalancePartitioner[PVEvent.Entity]))
and the constructor is :
/**
* Creates a FlinkKafkaProducer
Hi Kostas,
No, the same code was used.
I (1) started the job, (2) created a savepoint, (3) cancelled the job, (4)
restored the job with the same command as in (1) with the addition "-s
".
Regards,
Averell
--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
Hi,
what is the exact error message you are getting?
Best,
Dawid
On 10/10/18 11:51, 远远 wrote:
> invoke FlinkKafkaProducer011 constructor in scala:
> val producer = new
> FlinkKafkaProducer011[PVEvent.Entity](appConf.getPvEventTopic, new
> PvEventSerializeSchema, producerProps,
> Optional.of
IDEA only show "cannot resolve constructor" error message, without other
error message.
Dawid Wysakowicz 于2018年10月10日周三 下午5:55写道：
> Hi,
>
> what is the exact error message you are getting?
>
> Best,
>
> Dawid
>
> On 10/10/18 11:51, 远远 wrote:
>
> invoke FlinkKafkaProducer011 constructor in scala:
Hi,
We are using Flink 1.6.1 on yarn with rocksdb as backend incrementally
checkpointed to hdfs (for data and timers).
The job reads events from kafka (~1 billion event per day), constructs user
sessions using an EventTimeSessionWindow coupled with a late firing trigger
and WindowFunction with Ag
Hi Pawel,
As far as I know, the application attempt is incremented if the application
master fails and a new one is brought up. Therefore, what you are seeing
should not happen. I have just deployed on AWS EMR 5.17.0 (Hadoop 2.8.4) and
killed the container running the application master – the cont
Hi,
I would assume that the problem about blocked processing during a checkpoint is
caused by [1], because you mentioned the use of RocksDB incremental checkpoints
and it could be that you use it in combination with heap-based timers. This is
the one combination that currently still uses a sync
Hi Rinat,
Thanks for discussing this idea. Yes, I think this would be a good feature.
Can you open a Jira issue and describe the feature?
Thanks, Fabian
Am Do., 4. Okt. 2018 um 19:28 Uhr schrieb Rinat :
> Hi mates, in accordance with the contract of
> org.apache.flink.formats.avro.Deserializati
Hi Averell,
I'd go with approach 2). As of Flink 1.6.0 you can delete timers.
But even if you are on a pre-1.6 version, a ProcessFunction would be the
way to go, IMO.
You don't need to register a timer for each event.
Instead, you can register the first timer with the first event and have a
state
Hi,
what's the difference between partitions and subpartitions?
Thanks.
CM
Hi Fabian, I have created the issue,
https://issues.apache.org/jira/browse/FLINK-10525
Thx !
> On 10 Oct 2018, at 16:47, Fabian Hueske wrote:
>
> Hi Rinat,
>
> Thanks for discussing this idea. Yes, I think this would be a good feature.
> Can you open a Jira issue and describe the feature?
>
Hi Borys,
I remember that another user reported a similar issue recently [1] –
attached
to the ticket you can find his log file. If I recall correctly, we concluded
that YARN returned the containers very quickly. At the time, Flink's debug
level logs were inconclusive because we did not log the re
Hi everyone,
it seems Flink only supports global watermarks currently which is a problem for
my use case.
Many sensors send data which might be buffered for days in upstream systems
before arriving at the Flink job.
The job keys the stream by sensor. If other sensors send values in the
meantime
You are correct that watermarks are not tracked per key. You are dealing
with events with a high degree of delay variability. That is usually not a
good match for event time processing as implemented in Flink.
You could use event time processing and configure a very large window
allowed lateness
Hi,
We are facing an issue in standalone HA mode in Flink 1.4.0 where
Taskmanager restarts and is not able to register with the Jobmanager. It
times out awaiting *AcknowledgeRegistration/AlreadyRegistered* message from
Jobmanager Actor and keeps sending *RegisterTaskManager *message. The logs
at
Hi Xuefu,
Appreciate this proposal, and like Fabian, it would look better if you can
give more details of the plan.
Thanks, vino.
Fabian Hueske 于2018年10月10日周三 下午5:27写道：
> Hi Xuefu,
>
> Welcome to the Flink community and thanks for starting this discussion!
> Better Hive integration would be re
Hello,
I am have a job that reads messages from kafka, processes them and writes
back to kafka, this jobs works fine on flink 1.3.2. I upgraded cluster to
1.6.1 but now see below error. Has any one faced similar issue?
I have updated all the dependencies to use
1.6.1
org.apache.flink
f
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，
Partition is the output of a JobVertex which you can simply thought
contains an operator. And in real world, JobVertex will run in parallel,
each
will output some data, which is conceptually called subpartition.
Best,
Kurt
On Thu, Oct 11, 2018 at 10:27 AM Renjie Liu wrote:
> Hi, Chris:
>
Hi Fabian/Vno,
Thank you very much for your encouragement inquiry. Sorry that I didn't see
Fabian's email until I read Vino's response just now. (Somehow Fabian's went to
the spam folder.)
My proposal contains long-term and short-terms goals. Nevertheless, the effort
will focus on the followin
Hi Chandu,
What mode does your Flink run in?
In addition, can you check if the flink-metrics-core is included in the
classpath of the Flink runtime environment?
Thanks, vino.
Chandu Kempaiah 于2018年10月11日周四 上午9:51写道：
>
> Hello,
>
> I am have a job that reads messages from kafka, processes them
Hi，
I changed as below configurations，and it looks fine when job started.
But there’re no results issued when window ends after running about six hours,
and no errors and exceptions.
How can I position the question?
Yours,
September
发件人: 潘 功森
发送时间: Wednesday, Oc
Hi,
I'm trying to run an integration test of my flink application. My test code
looks like this:
public class HttpsCsvIngestorTest extends AbstractTestBase {
private final static Logger LOG =
LoggerFactory.getLogger(HttpsCsvIngestorTest.class);
@Test
public void testHttpsCsvIngestio
Would it maybe make sense to provide Flink as an engine on Hive
(„flink-on-Hive“)? Eg to address 4,5,6,8,9,10. this could be more loosely
coupled than integrating hive in all possible flink core modules and thus
introducing a very tight dependency to Hive in the core.
1,2,3 could be achieved via
Flink is running as standalone cluster in High Availability mode, My
application jar is a fat jar which has all the necessary dependencies
included.
I will check once again and verify by adding the flink-metrics-core to the
classpath.
Thanks
Chandu
On Wed, Oct 10, 2018 at 8:38 PM vino yang wrot
I'm using flink 1.5.0. The test gives the same error even with flink-1.6.0.
Also, I introduced a Thread.sleep(3); before the assert statement. That
didn't help either.
Regards,
James
On Thu, Oct 11, 2018 at 11:11 AM James Isaac
wrote:
> Hi,
>
> I'm trying to run an integration test of my fl
Hi gongsen,
Have you used event time as time semantics? If so, then the possible
problem is related to watermark.
Since I don't know the details of your program, it's hard to make a
conclusion. You can check if your watermark is firing normally.
Thanks, vino.
潘 功森 于2018年10月11日周四 下午12:12写道：
>
43 matches
Mail list logo
