ory. If
>>> the
>>> > setting is too small, the task cannot be started. If the setting is too
>>> > large, there may be a waste of resources. As far as possible, Flink
>>> > framework can automatically set a reasonable value, but I have a small
>>> >
nly related to the parallelism of the
>> task,
>> > but also to the complexity of the task DAG. The more complex a DAG is,
>> > shuffle write and shuffle read require larger buffers. How can we
>> determine
>> > how many RS and IG a DAG has?
>> >
>>
rmine
> > how many RS and IG a DAG has?
> >
> >
> >
> > Best
> > JasonLee
> >
> >
> > Replied Message
> > | From | Yuxin Tan |
> > | Date | 12/28/2022 18:29 |
> > | To | |
> > | Subject | Re: [DISCUSS]
value, but I have a small
> > problem. network memory is not only related to the parallelism of the
> task,
> > but also to the complexity of the task DAG. The more complex a DAG is,
> > shuffle write and shuffle read require larger buffers. How can we
> determine
> > how ma
to the complexity of the task DAG. The more complex a DAG is,
> shuffle write and shuffle read require larger buffers. How can we determine
> how many RS and IG a DAG has?
>
>
>
> Best
> JasonLee
>
>
> Replied Message ----
> | From | Yuxin Tan |
> | Date |
uffle write and shuffle read require larger buffers. How can we determine
> how many RS and IG a DAG has?
>
>
>
> Best
> JasonLee
>
>
> Replied Message
> | From | Yuxin Tan |
> | Date | 12/28/2022 18:29 |
> | To | |
> | Subject | Re: [DISCUSS] F
a
DAG has?
Best
JasonLee
Replied Message
| From | Yuxin Tan |
| Date | 12/28/2022 18:29 |
| To | |
| Subject | Re: [DISCUSS] FLIP-266: Simplify network memory configurations for
TaskManager |
Hi, Roman
Thanks for the replay.
ExclusiveBuffersPerChannel and FloatingBuffersPerGate
Hi, Roman
Thanks for the replay.
ExclusiveBuffersPerChannel and FloatingBuffersPerGate are obtained from
configurations, which are not calculated. I have described them in the FLIP
motivation section.
> 3. Each gate requires at least one buffer...
The timeout exception occurs when the
Hi everyone,
Thanks for the proposal and the discussion.
I couldn't find much details on how exactly the values of
ExclusiveBuffersPerChannel and FloatingBuffersPerGate are calculated.
I guess that
- the threshold evaluation is done on JM
- floating buffers calculation is done on TM based on the
Hi, Weihua
Thanks for your suggestions.
> 1. How about reducing ExclusiveBuffersPerChannel to 1 first when the
total buffer is not enough?
I think it's a good idea. Will try and check the results in PoC. Before all
read buffers use floating buffers, I will try to use
(ExclusiveBuffersPerChannel
Hi Yuxin,
Thanks for the proposal.
"Insufficient number of network buffers" exceptions also bother us. It's
too hard for users to figure out
how much network buffer they really need. It relates to partitioner type,
parallelism, slots per taskmanager.
Since streaming jobs are our primary
Hi, all
Thanks for the reply and feedback for everyone!
After combining everyone's comments, the main concerns, and corresponding
adjustments are as follows.
@Guowei Ma, Thanks for your feedback.
> should we introduce a _new_ non-orthogonal
Hi Yuxin,
Thanks for creating this FLIP.
It's good if Flink does not require users to set a very large network
memory, or tune the advanced(hard-to-understand) per-channel/per-gate
buffer configs, to avoid "Insufficient number of network buffers" exceptions
which can easily happen for large
Hi Yuxin,
Thanks for the proposal!
After reading the FLIP, I have some questions about the default value.
This FLIP seems to introduce a *new* config
option(taskmanager.memory.network.required-buffer-per-gate.max) to control
the network memory usage.
1. Is this configuration at the job level or
Hi Yuxin,
Thanks for proposing the FLIP!
The motivation section makes sense. But it seems that the proposed change
section mixes the proposed config with the evaluation results. It is a bit
hard to understand what configs are proposed and how to describe these
configs to users. Given that the
If this improves the performance+১
On Sat, 24 Dec, 2022, 5:47 pm Guowei Ma, wrote:
> Hi,
> Thank you very much for driving this FLIP in order to improve user
> usability.
>
> I understand that a key goal of this FLIP is to adjust the memory
> requirements of shuffle to a more reasonable range.
Hi,
Thank you very much for driving this FLIP in order to improve user
usability.
I understand that a key goal of this FLIP is to adjust the memory
requirements of shuffle to a more reasonable range. Through this adaptive
range adjustment, the memory efficiency can be improved under the premise
Hi,
Thanks for driving this FLIP, +1 for the proposed changes.
Limit the maximum value of shuffle read memory is very useful when using
when using adaptive batch scheduler. Currently, the adaptive batch
scheduler may cause a large number of input channels in a certain TM, so we
generally
Thanks for the proposal, Yuxin.
+1 for the proposed changes. I think these are indeed helpful usability
improvements.
Best,
Xintong
On Mon, Dec 19, 2022 at 3:36 PM Yuxin Tan wrote:
> Hi, devs,
>
> I'd like to start a discussion about FLIP-266: Simplify network memory
> configurations for
Hi, devs,
I'd like to start a discussion about FLIP-266: Simplify network memory
configurations for TaskManager[1].
When using Flink, users may encounter the following issues that affect
usability.
1. The job may fail with an "Insufficient number of network buffers"
exception.
2. Flink network
20 matches
Mail list logo