退订
| |
lizikunn
|
|
lizik...@163.com
|
签名由网易邮箱大师定制
ELLEX_SHEN created FLINK-22883:
--
Summary: Select view columns fail when store metadata with hive
Key: FLINK-22883
URL: https://issues.apache.org/jira/browse/FLINK-22883
Project: Flink
Issue
I understand your scenario but I disagree with its assumptions:
"However, the partition of A is empty and thus A is temporarily idle." -
you're assuming that the behavior of the source is to mark itself idle if
data isn't available, but that's clearly source-specific and not behavior
we expect to
At least one big motivation is having (temporary) empty partitions. Let me
give you an example, why imho idleness is only approximate in this case:
Assume you have source subtask A, B, C that correspond to 3 source
partitions and a downstream keyed window operator W.
W would usually trigger on
Yes I'm talking about an implementation of idleness that is unrelated to
processing time. The clear example is partition assignment to subtasks,
which probably motivated Flink's idleness functionality in the first place.
On Fri, Jun 4, 2021 at 12:53 PM Arvid Heise wrote:
> Hi Eron,
>
> Are you
Hi Eron,
Are you referring to an implementation of idleness that does not rely on a
wall clock but on some clock baked into the partition information of the
source system?
If so, you are right that it invalidates my points.
Do you have an example on where this is used?
With a wall clock, you
Little update on this, more good discussion over the last few days, and the
FLIP will probably be amended to incorporate idleness. Voting will remain
open until, let's say, mid-next week.
On Thu, Jun 3, 2021 at 8:00 AM Piotr Nowojski wrote:
> I would like to ask you to hold on with counting
I agree with Steven. This logic can be added in a dynamic config framework
that can bind into Flink operators. We probably don't need to let Flink
runtime handle it.
On Fri, Jun 4, 2021 at 8:11 AM Steven Wu wrote:
> I am not sure if we should solve this problem in Flink. This is more like
> a
Dawid, I think you're mischaracterizing the idleness signal as inherently a
heuristic, but Flink does not impose that. A source-based watermark (and
corresponding idleness signal) may well be entirely data-driven, entirely
deterministic. Basically you're underselling what the pipeline is capable
Piotr Nowojski created FLINK-22882:
--
Summary: Tasks are blocked while emitting records
Key: FLINK-22882
URL: https://issues.apache.org/jira/browse/FLINK-22882
Project: Flink
Issue Type: Bug
Piotr Nowojski created FLINK-22881:
--
Summary: Tasks are blocked while emitting stream status
Key: FLINK-22881
URL: https://issues.apache.org/jira/browse/FLINK-22881
Project: Flink
Issue
While everything I wrote before is still valid, upon further rethinking, I
think that the conclusion is not necessarily correct:
- If the user wants to have pipeline A and B behaving as if A+B was jointly
executed in the same pipeline without the intermediate Pulsar topic, having
the idleness in
I am not sure if we should solve this problem in Flink. This is more like a
dynamic config problem that probably should be solved by some configuration
framework. Here is one post from google search:
Hi everyone,
Flink jobs are always long-running. When the job is running, users
may want to control the job but not stop it. The control reasons can be
different as following:
1.
Change data processing’ logic, such as filter condition.
2.
Send trigger events to make the
I think the core issue in this discussion is that we kind of assume that
idleness is something universally well-defined. But it's not. It's a
heuristic to advance data processing in event time where we would lack data
to do so otherwise.
Keep in mind that idleness has no real definition in terms
As I've said I am not a security expert and that's why I have to ask for
clarification, Gabor. You are saying that if we configure a truststore for
the REST endpoint with a single trusted certificate which has been
generated by the operator of the Flink cluster, then the attacker can
generate a
Timo Walther created FLINK-22880:
Summary: Remove "blink" term in code base
Key: FLINK-22880
URL: https://issues.apache.org/jira/browse/FLINK-22880
Project: Flink
Issue Type: Sub-task
Timo Walther created FLINK-22879:
Summary: Remove "blink" suffix from table modules
Key: FLINK-22879
URL: https://issues.apache.org/jira/browse/FLINK-22879
Project: Flink
Issue Type:
Hi,
> Imagine you're starting consuming from the result channel in a situation
were you have:
> record4, record3, StreamStatus.ACTIVE, StreamStatus.IDLE record2,
record1, record0
> Switching to the encoded StreamStatus.IDLE is unnecessary, and might
cause the record3 and record4 to be late
Timo Walther created FLINK-22878:
Summary: Allow placeholder options in format factories
Key: FLINK-22878
URL: https://issues.apache.org/jira/browse/FLINK-22878
Project: Flink
Issue Type:
Timo Walther created FLINK-22877:
Summary: Remove BatchTableEnvironment and related API classes
Key: FLINK-22877
URL: https://issues.apache.org/jira/browse/FLINK-22877
Project: Flink
Issue
Hi all,
@Xintong Song
Thanks for reminding me, I would contact Jark to update the wiki page.
Besides, I'd like to provide more inputs by sharing our experience about
upgrading Internal version of Flink.
Flink has been widely used in the production environment since 2018 in our
company. Our
Arvid Heise created FLINK-22876:
---
Summary: Adding SharedObjects junit rule to ease test development
Key: FLINK-22876
URL: https://issues.apache.org/jira/browse/FLINK-22876
Project: Flink
Issue
JYXL created FLINK-22875:
Summary: Cannot work by Pickler and Unpickler
Key: FLINK-22875
URL: https://issues.apache.org/jira/browse/FLINK-22875
Project: Flink
Issue Type: Bug
Reporter:
Hi,
Thanks for picking up this discussion. For the record, I also think we
shouldn't expose latency markers.
About the stream status
> Persisting the StreamStatus
I don't agree with the view that sinks are "storing" the data/idleness
status. This nomenclature makes only sense if we are
Spongebob created FLINK-22874:
-
Summary: flink table partition trigger doesn't effect as
expectation when sink into hive table
Key: FLINK-22874
URL: https://issues.apache.org/jira/browse/FLINK-22874
> I did not mean for the user to sign its own certificates but for the
operator of the cluster. Once the user request hits the proxy, it should no
longer be under his control. I think I do not fully understand yet why this
would not work.
I said it's not solving the authentication problem over any
I believe that the correctness of watermarks and stream status markers is
determined entirely by the source (ignoring the generic assigner). Such
stream elements are known not to overtake records, and aren't transient
from a pipeline perspective. I do agree that recoveries may be lossy if
some
Chesnay Schepler created FLINK-22873:
Summary: Add ToC to configuration documentation
Key: FLINK-22873
URL: https://issues.apache.org/jira/browse/FLINK-22873
Project: Flink
Issue Type:
Hi all,
Very thanks @Dawid for resuming the discussion and very thanks @Till for the
summary ! (and very sorry for I missed the mail and do not response in time...)
I also agree with that we could consider the global commits latter separately
after we have addressed the final checkpoints, and
I did not mean for the user to sign its own certificates but for the
operator of the cluster. Once the user request hits the proxy, it should no
longer be under his control. I think I do not fully understand yet why this
would not work.
What I would like to avoid is to add more complexity into
Hi!
I think there might be possible alternatives but it seems Kerberos on the
rest endpoint ticks all the right boxes and provides a super clean and
simple solution for strong authentication.
I wouldn’t even consider sidecar proxies etc if we can solve it in such a
simple way as proposed by G.
Till, thanks for investing time in giving further options.
Marci, thanks for summarizing the use-case point of view.
We've arrived back to one of the original problems. Namely if an attacker
gets access to a node it's possible to cancel other user's jobs (and more
can be done).
Self signed
I am not saying that we shouldn't add a strong authentication mechanism if
there are good reasons for it. I primarily would like to understand the
context a bit better in order to give qualified feedback and come to a good
decision. In order to do this, I have the feeling that we haven't fully
Hi Eron,
I might be missing some background on Pulsar partitioning but something
seems off to me. What is the chunk/batch/partition that Pulsar brokers
will additionally combine watermarks for? Isn't it the case that only a
single Flink sub-task would write to such a chunk and thus will produce
Hi all,
We are using Flink for our eventing system. Overall we are very happy with
the tech, documentation and community support and quick replies in mails.
My last 1 year experience with versions.
We were working on 1.10 initially during our research phase then we
stabilised with 1.11 as we
36 matches
Mail list logo