Thanks. The plan is to use the DataStream API (or possibly Beam) which is fine 
for data from Kafka (and presumably from a file too). Assuming that I don't 
stop and restart the streaming job but just leave it idle (as no events are 
coming into Kafka), is there any issue with leaving a window 'open' for a few 
hours? The window would be defined as 5 minutes of event time but due to the 
fact that the upstream system is buffering the data Flink would only see the 
end of those 5 minutes after many hours (processing time).

I don't have specific details of what the Flink job will be doing because I 
want this architecture to support many usecases of processing the data – I'm 
trying to determine if it makes sense to model the data as a stream, coming via 
Kafka despite the Kafka stream being populated in large chunks (that aren't to 
do with event time, but rather due to batch-based processing of the upstream 
process) or not. We would be typically using a keyed sliding window. The 
watermark of the window would be defined in terms of event time. The event time 
would not be progressing for the extended period between the end of one 
upstream batch and the next so the windows would remain open for that time.

Thanks,
Jonny
From: miki haiat [mailto:[email protected]]
Sent: Monday, January 21, 2019 5:07 PM
To: Jonny Graham <[email protected]>
Cc: [email protected]
Subject: Re: Kafka stream fed in batches throughout the day

In flink you cant read data from kafka in Dataset API (Batch)
And you dont want to mess with start and stop your job every few hours.
Can you elaborate more on your use case ,
Are you going to use KeyBy , is thire any way to use trigger ... ?



On Mon, Jan 21, 2019 at 4:43 PM Jonny Graham 
<[email protected]<mailto:[email protected]>> wrote:
We have a Kafka stream of events that we want to process with a Flink 
datastream process. However, the stream is populated by an upstream batch 
process that only executes every few hours. So the stream has very 'bursty' 
behaviour. We need a window based on event time to await the next events for 
the same key. Due to this batch population of the stream, these windows can 
remain open (with no event activity on the stream) for many hours. From what I 
understand we could indeed leave the Flink datastream process up and running 
all this time and the window would remain open. We could even use a savepoint 
and then stop the process and restart it (with the window state being restored) 
when we get the next batch and the events start appearing in the stream again.

One rationale for this mode of operation is that we have a future usecase where 
this stream will be populated in real-time and would behave like a normal 
stream.

Is that a best-practice approach for this scenario? Or should we be treating 
these batches as individual batches (Flink job that ends with the end of the 
batch) and manually handle the windowing that needs to cross multiple batches.

Thanks,
Jonny

Confidentiality: This communication and any attachments are intended for the 
above-named persons only and may be confidential and/or legally privileged. Any 
opinions expressed in this communication are not necessarily those of NICE 
Actimize. If this communication has come to you in error you must take no 
action based on it, nor must you copy or show it to anyone; please 
delete/destroy and inform the sender by e-mail immediately.
Monitoring: NICE Actimize may monitor incoming and outgoing e-mails.
Viruses: Although we have taken steps toward ensuring that this e-mail and 
attachments are free from any virus, we advise that in keeping with good 
computing practice the recipient should ensure they are actually virus free.

Confidentiality: This communication and any attachments are intended for the 
above-named persons only and may be confidential and/or legally privileged. Any 
opinions expressed in this communication are not necessarily those of NICE 
Actimize. If this communication has come to you in error you must take no 
action based on it, nor must you copy or show it to anyone; please 
delete/destroy and inform the sender by e-mail immediately.
Monitoring: NICE Actimize may monitor incoming and outgoing e-mails.
Viruses: Although we have taken steps toward ensuring that this e-mail and 
attachments are free from any virus, we advise that in keeping with good 
computing practice the recipient should ensure they are actually virus free.

Confidentiality: This communication and any attachments are intended for the 
above-named persons only and may be confidential and/or legally privileged. Any 
opinions expressed in this communication are not necessarily those of NICE 
Actimize. If this communication has come to you in error you must take no 
action based on it, nor must you copy or show it to anyone; please 
delete/destroy and inform the sender by e-mail immediately.
Monitoring: NICE Actimize may monitor incoming and outgoing e-mails.
Viruses: Although we have taken steps toward ensuring that this e-mail and 
attachments are free from any virus, we advise that in keeping with good 
computing practice the recipient should ensure they are actually virus free.

Confidentiality: This communication and any attachments are intended for the 
above-named persons only and may be confidential and/or legally privileged. Any 
opinions expressed in this communication are not necessarily those of NICE 
Actimize. If this communication has come to you in error you must take no 
action based on it, nor must you copy or show it to anyone; please 
delete/destroy and inform the sender by e-mail immediately.  
Monitoring: NICE Actimize may monitor incoming and outgoing e-mails.
Viruses: Although we have taken steps toward ensuring that this e-mail and 
attachments are free from any virus, we advise that in keeping with good 
computing practice the recipient should ensure they are actually virus free.

Reply via email to