Just want to ask, according to fluss cluster architecture on 1:28:08, Does Fluss containes paimon or iceberg within its source code? (Like Flink CDC uses Debezium) or those two table formats just pluggable elements, working independently with fluss itself? Also, If Fluss uses arrow, Is C++ a main language for fluss? (I guess Rust maybe...)

It's super hard to follow. Such a pity because it's a great explanation of pyflink sdk.

Holographic representation rocks!

robot talks on Flink

Nesito yo estoy en mexico no puedo salir demexico canselaron los biajes 29 ya estoy aya sanfrasico undiscupa

Buenas noche cuando kieres aser bisnes los utimo deste mes pliz cuando

Great insights

the accent!!!!!!!!!!!!

I am confused. How is this different from Kafka Streams and KTable?

If you have multiple fraud-counter http servers or multiple merchant-scorer http servers running (for high-availability), does the storage synced across those servers? If so, how does the syncing works?

The question about partial deserializing wasn't bad. Interesting.

where can I find the link of demo at the description? @8:38

Very prescient talk

I'll build on this

Good talk but the audio is not that great.

no wonder 781 people watched this in last 3 years....

Very useful introduction of Apache Flink. I was able to test Flink features using Google Colab.

Hi, is there a way to improve the subtitles ? It's very hard to understand the audio and subtitles don't help at all.

What's the disadvantage of solving this problem with Kafka stream?

운전점수 확인 정말 유용하게 사용 중인데 엄청난 기술력이 있었네요.

cool

Please an update of pyflink processing kafka topics

Hi, that's really interesting! I have 2 questions: How do you handle race conditions? In the example, if the counter and merchant functons respond at the same time, they could both think they're the first to answer and we end up with both in cache but not sent to the model, or does the context run one message at a time guarantied, like in Akka? How do you handle resilience? THe 30 days delay on a message is a great feature made simple here, but let say the app dies, and is rebooted, is it going to be lost? Is there a sync to something like S3 for the state?

I guess from 30:00 onwards Flink have addressed this with HybridSource? Havent used it yet but seen a few talks on it

The second guy is brilliant

Noice.

Thank you for sharing the video. However, it seems that Flink is not performing well in terms of application upgrade. For instance, if we have a lambda that consumes Kinesis stream, a lambda update won't result in any downtime or duplicate records and everything will be handled smoothly. On the other hand, Flink application requires us to handle several aspects such as save point, checkpoint, blue-green deployment, dual stack, and duplicate data on our own. Even Amazon Managed Flink and KDA may experience downtime during application upgrade. I am optimistic that Flink can do better than current one

This is amazing, we have a similar usecase in our company and i was confused between feast, sagemaker. I was also worried about explaining the infrastructure requirements to PMs, managers if we chose Feast. This has simplified the whole setup and ideal usecases for small startups.

Hello, thank you for the video

Whether there is need to install RocksDB in kubernates cluster or just adding state.backend: rocksdb ... these line in conf/flink_conf.yaml is enough?

Hi, thanks for course. I have a question about first demo, there is no output when "table.limit(10).execute().print()" code runned. Have you got any iade about this problem ?

Great Presentation!! Neatly sequenced with how the problems were conquered and the way forward.

What kind of join do you propose between 2 streams if there are more than 1 key possible. For the example you explained suppose one needs to join 2 streams, order and sale, either on sale_id or userid.

Great video! Thanks

Can you please create a sample project on this..

How does Flink CEP Manages it state? We take patterns in real time.... but flink does not removes pattern from state whose window time is over?.. is there any solution

0.75x speed should be a lot better

This is just a brilliant piece of information!

How does unordered input cause checkpoint failure. Can you please elaborate on the use case which caused this failure. My understanding is that even if you add allowed lateness to consider unordered event, even then checkpoint will keep happening to store current state.

does anyone know where I can find the state management processes that TableAPI uses, I understand that is under the hood and is being done automatically, but was hoping to understand further how it does so, such that I could use it in my own project.

How does the flink application in blue environment catch-up with the lag introduced due to re-deployment ?

Thank you for the video. A quick honest feedback. Presentation is disorienting. I would like to hear it like: * What Flink is * Who uses it? why? (not generic, specific) * What you are introducing in this video, (story, expectation) * Demo * show requirements * show the solution * show what's implemented * explain what you do the things you do * Conclude * What if you don't use flink (flink vs hard way to do it)

Very useful and there is not such less learned content about Flink implementation sharing. Wandering about the new version of Flink will resolve any of those issues yet ?

only keep last 2 checkpoints . Is is configurable for incremental checkpoint ??

It feels not correct that the optionally scaling apply to any of the "Aggregation" and "Windowed Retention" as the state size is promotional to the number of keys which may varies depending on the input. If we say the group-by keys space are finite and we have an estimate for that, then it can fit in optionally scaling.

I installed flink but can not access the UI.

great talk

Hello, this is excellent video. Very helpful. I have e requirements to insert data into Parent and Child table, can you please suggest or share a reference o how to do this?

Flink Forward is becoming my favourite conference of all. To the point, tech stuff and no jargon’s. Thank you Brent for all the valuable points.

This was really informative, appreciate the troubleshooting scenarios. Cheers David!