This is a gem of a video series with an even greater gem of a presenter. I truly mean this when I say that Tim is just insanely good at breaking down info like this. He deserves all the promotions and love that can be humanely offered by any corporation on this planet. I encourage him to keep making videos like this on more topics. My sole regret is that this video did not come out years earlier. Tim, if you're reading this, I applaud you. You are unironically my hero of this month and I hope you continue living a great life.
official and legit man! I just wondering why this kind of official tech videos are not listed at the top when searching "what is kafka". bunch of superficial and low quality videos show up and contains wrong information and non-core details. really hope you guys PR these videos to let more engineers get the correct stuffs.
This speaker is great, with a casual style and different intonation that's easy to follow. In a lot of tech RUclips videos the speaker just rambles on in a monotone voice making me want to zone out.
Well Explained. Question : How the Disk Space underneath the Brokers/Segments Grow? Is that something Producers or Consumers need to be worried about? That's Cloud Offering as a SaaS or IaaS?
Done thanks Logs are immutable, can have multiple consumers consuming from different offsets of the log because consuming doesn’t actually delete the messages from the log Partitions of a topic live on different brokers and not all brokers must have partitions from a topic Partitions are replicated across brokers, so that if broker fails the topic partition is replicated elsewhere and can change the replication factors for partitions. Master/leader for each partition when it’s replicated and writes happen at the leader CLI producer can be used for testing How does a producer know which partition to write the message to? Partitioning strategy by default uses the hash(key) % numpartitions Messages with the same key will land in the same partition and will maintain their ordering. (In the case that number of partitions in a topic changes then this isn’t the case anymore but they shouldn’t change) Each consumer has an offset of where in the log it’s reading from. Consumer pulls messages from topics after offset n
10:55 Cornflower blue, you say, CORNFLOWER BLUE ??? I'm shocked. SHOCKED to find out that colors aren't being identified in here" Why, that is clearly a case of #7cb0f9, which as anyone know is soft blue! Thank you for the wonderful video. You, sir rock. _List below is for me to quickly find stuff. Be warned, these time points don't do justice. Watch the whole video. It's awesome_ 0:02 - Nice little joke here 0:09 - Nice upbeat music. Don't miss this one! 0:20 - Synopsis of what you will learn in 24 minutes 0:54 - Kafka's job 0:59 - What's a producer 1:33 - Data stuff that goes into a Kafka Cluster 1:57 - What's in a Kafka Cluster ? 2:21 - Brokers 3:32 - Consumers 3:57 - Relationship Consumer and Application 4:15 - Reiteration of Fundamental parts of Kafka 4:52 - ZooKeeper 5:45 - Decoupling of Producers and Consumers 6:51 - What does ZooKeeper really do in a Kafka Cluster ? 7:39 - Topics 9:11 - Partitions 10:19 - Segments 11:01 - Topology and Detail example of a Kafka Cluster 12:25 - What's a Log 14:02 - Consumers. Do they really consume ? 14:55 - Structure of a Kafka Message 16:10 - Brokers revisited 17:19 - Broker Replication 18:16 - Producers revisited 19:51 - Load Balancing 21:25 - Consumers revisited 22:41 - Distributed Consumption
Thanks Tim for the awesome explanation of Kafka terms and how they relate with each other. After watching this video only I could understand kafka terms in real deep and I no more have to cram these terms again amd again.🙂
When he mentioned timestamps, I was imagining a cop saying “You have the right to a time stamp. If you cannot afford a time stamp, one will be provided for you.”
8:50 One slip here. I think it should be "the broker, that topic lives on" and not "that partition lives on", because 1) you haven't even introduced partitions yet, at this very moment, and 2) you later say "if you're writing messages into that topic".. which again confirms my point, that you mean topic, not a partition.
why kafka use consumer and suscriber in their terminology ? this are different approach in messaging , so what Kafka model use ? consumer/producer or publish/suscriber ?
Hi, I am fresher, and new to Kafka. For storing those topics we need persistent storage and consuming a message don't delete that message. Now my question is suppose I have X ammount of persistent storage and producer produces X ammount message in Y days, what will happen to my storage after Y'th day?
Thanks for this great video! I have a few questions that hopefully someone could help clarify :) 1. How are brokers replicated? It sounds like it's async replicated, hence I imagine when the leader failover there would be some small amount of msg loss (because replica would always lag primary a tiny bit)? 2. When the broker receives the msg, does it write to log immediately or does it do some kind of in-memory buffering and write by small batch? And if so what happens to the non-flushed messages if that broker crashes? Just to clarify, I'm not criticizing Kafka, it is a great tool and I really liked it while working with it in my previous job. But I'm just curious because I've heard various techtalks about how kafka is used in various products -- e.g. nu bank which is a financial startup uses kafka and from their talk I had this feeling that they rely on kafka for being 100% reliable (as in, not losing messages), which surprises me. 3. One last noob question, apparently Kafka's great capability for supporting high write throughput is partially due to its sequential write, hence avoiding random disk seek. But given that consumer almost always consumes the message with a slight delay, does it mean whenever a consumer pulls new messages it breaks this nice sequential mechanism (because we need to seek to a different disk location than the end of the log file)? Thanks!
To answer question #1 (after watching the next video in the series), apparently one can tune the replication iin Kafka to only have producer acked after the msg is replicated to all replicas.
Hey have you found the answers to your questions #2 & #3? I don't think any system can be 100% reliable but considering industry wide adoption I believe they must be doing something really well. I would love to know their solution to these fundamental data problems
At 4:53 why you have `n` everywhere? That might be confusing since it is a kind of implication that we have equal number of components on all the levels.
This guy is the best tech educator I've ever seen.
This is a gem of a video series with an even greater gem of a presenter. I truly mean this when I say that Tim is just insanely good at breaking down info like this. He deserves all the promotions and love that can be humanely offered by any corporation on this planet. I encourage him to keep making videos like this on more topics. My sole regret is that this video did not come out years earlier. Tim, if you're reading this, I applaud you. You are unironically my hero of this month and I hope you continue living a great life.
Yeah, Tim's very charismatic, approachable and well-spoken. Most people aren't. Respect.
I felt same way and came here to write same . Saw your message so just "Reused". Loud claps for Tim.
I fell asleep on most other technology videos but Tim is exceptional good. His skill on verbal fluency and the use of context keeps me wide awake.
absolutely agreee!
Tim you’re the best
official and legit man! I just wondering why this kind of official tech videos are not listed at the top when searching "what is kafka". bunch of superficial and low quality videos show up and contains wrong information and non-core details. really hope you guys PR these videos to let more engineers get the correct stuffs.
The Best Kafka fundamentals video. Period. Give the instructor an award. So precise and to the point.
This speaker is great, with a casual style and different intonation that's easy to follow. In a lot of tech RUclips videos the speaker just rambles on in a monotone voice making me want to zone out.
Straight forwared, right on point, no bullshit. Cleared all main kafka concepts in just 24 mins.
Nobody ever taught Kafka in details like you did Tim. Much appreciated 👍🏻
The best Kafka tutorial I have ever seen !
i can state unequivocally that i have never come across a better video instructor.. Hats off
I can't believe I just watch a 24 minute technical video without yawning or pausing. This guy is good! 🏆🏆
That was one of the best description of Apache Kafka fundamentals that I saw, thanks
This guy’s presentation skills should be made gold standard, period !!!
I can say that I am not watching any other channel for Kafka now.good job!!!
I am loving this guy! He is so to the point, makes joke sometimes and he explains it quite simply. I just like these vidoes. Awesome!!!!
Great that the day has come that Zookeeper is no longer required - nice that you were able to give a heads up!
Such clarity of thoughts and impeccable explanation, answered all the questions that came to my mind. Best Kafka video!
Thank you, Tim and the Confluent team, your tutorials are top-notch and really help people understand the subject. Wish you all the best!
Amazing video.
Very thorough.
Can't believe this is free. Thanks Confluent.
Much appreciated! Thanks.
Highly recommended for freshers with couple of years of experience, starting journey towards resiliency and kafka.
Halfway through the video and already loving this guy. Super presentation and delivery skills. Kudos Tim!!
One of the best videos on Kafka basics and understanding of the Cluster
Amazing way of explaining and great choice and structure of sentences to get the concept across effectively. Great job
this is exactly i was looking for.. best of all the videos on Kafka Fundamentals
First time learning about Kafka, learned everything I needed to know from your one video. You are a great presenter sir, thank you!
Tims explanation!...top notch !
Wow this is easily understandable and not boring. Is it too much to ask to make videos of this quality of other topics in tech..?
Who is this guy ? Blew my mind away
Man your rocked this presentation! Thank you Tim.
Hi Tim, thank you for a great explanation. Also thumbs up to the team behind the video.
Excellent content 👍🏽 to the point, no fluff, clearly explained with diagrams 👍🏽
Dear Confluent,
Can we have Tim do videos on Spark, Druid and Kubernetes too?
Thank you for the video. Helps me a lot to understand the components of Kakfa.
Great explanation with good detail. thankyou for your great effort.
Awesome presentation, crystal clear
AMazing videos. Consider using patterns or other visual indicators in place of color to accommodate various color-blindnesses
Amazing.. crystal clear explanation Tim.. Thanks a lot ..
Explanation is gorgeous !
Very well compiled, i struggled a lot and you have put all information at one place! cheers!
Well Explained.
Question : How the Disk Space underneath the Brokers/Segments Grow? Is that something Producers or Consumers need to be worried about? That's Cloud Offering as a SaaS or IaaS?
Very well explained. My only gripe is that I see that more as an azure blue
Excelent tutorial! Straight to the point and super clear
Entering now in Kafka and finally a perfect simple rapid nice explanation. Thank you for this video!
Done thanks
Logs are immutable, can have multiple consumers consuming from different offsets of the log because consuming doesn’t actually delete the messages from the log
Partitions of a topic live on different brokers and not all brokers must have partitions from a topic
Partitions are replicated across brokers, so that if broker fails the topic partition is replicated elsewhere and can change the replication factors for partitions. Master/leader for each partition when it’s replicated and writes happen at the leader
CLI producer can be used for testing
How does a producer know which partition to write the message to? Partitioning strategy by default uses the hash(key) % numpartitions
Messages with the same key will land in the same partition and will maintain their ordering. (In the case that number of partitions in a topic changes then this isn’t the case anymore but they shouldn’t change)
Each consumer has an offset of where in the log it’s reading from. Consumer pulls messages from topics after offset n
Nice intro! Other complicated software should take note of this simple introductionairy video
Actually such a great explanation.
Very good presenter and very informative video, thanks for sharing! 😄
Such great content and so much detail oriented that it's cleared my all doubts. Thanks for making such content. 👍
Tim, You are very good explainer bro
Excellent explanation in a short video Tim. Really appreciate your effort. Thank you!!
Just wanted to express my appreciation for this (university level) lecture! Thank you, very well done!
Top Notch explanation/video!! Well done!
its a great video to understand the basics, thanks for posting this
Awesome, Very simple to understand and managed to hold the attention
10:55 Cornflower blue, you say, CORNFLOWER BLUE ??? I'm shocked. SHOCKED to find out that colors aren't being identified in here" Why, that is clearly a case of #7cb0f9, which as anyone know is soft blue! Thank you for the wonderful video. You, sir rock.
_List below is for me to quickly find stuff. Be warned, these time points don't do justice. Watch the whole video. It's awesome_
0:02 - Nice little joke here
0:09 - Nice upbeat music. Don't miss this one!
0:20 - Synopsis of what you will learn in 24 minutes
0:54 - Kafka's job
0:59 - What's a producer
1:33 - Data stuff that goes into a Kafka Cluster
1:57 - What's in a Kafka Cluster ?
2:21 - Brokers
3:32 - Consumers
3:57 - Relationship Consumer and Application
4:15 - Reiteration of Fundamental parts of Kafka
4:52 - ZooKeeper
5:45 - Decoupling of Producers and Consumers
6:51 - What does ZooKeeper really do in a Kafka Cluster ?
7:39 - Topics
9:11 - Partitions
10:19 - Segments
11:01 - Topology and Detail example of a Kafka Cluster
12:25 - What's a Log
14:02 - Consumers. Do they really consume ?
14:55 - Structure of a Kafka Message
16:10 - Brokers revisited
17:19 - Broker Replication
18:16 - Producers revisited
19:51 - Load Balancing
21:25 - Consumers revisited
22:41 - Distributed Consumption
Hey :) Your presentation and explanatory style is really great - excellent in fact, and just at the right level for me! Many thanks
Thanks Tim for the awesome explanation of Kafka terms and how they relate with each other. After watching this video only I could understand kafka terms in real deep and I no more have to cram these terms again amd again.🙂
Cool, Unbelievable such a valuable information is for free.
Thanks Tim for this excellent video.
This was awesome! top-notch! Bravo
Great video! Simple to understand and managed to hold attention
fantastic video! BTW I like the shirt 😎
Awesome explanation. Thank you
Loved it. Thanks for making it easy to understand.
Excellent explanation. Thank you
This guy PRESENTS.
When he mentioned timestamps, I was imagining a cop saying “You have the right to a time stamp. If you cannot afford a time stamp, one will be provided for you.”
Solid knowledge, thank you fro sharing this.
well explained thank you Sir
8:50 One slip here. I think it should be "the broker, that topic lives on" and not "that partition lives on", because 1) you haven't even introduced partitions yet, at this very moment, and 2) you later say "if you're writing messages into that topic".. which again confirms my point, that you mean topic, not a partition.
Great video, thanks alot!
great video and even better series!
Super course, very effective and well delivered. Thanks !
Thanks great breakdown and presentation .
Very well explained. Thank You so much :)
why kafka use consumer and suscriber in their terminology ? this are different approach in messaging , so what Kafka model use ? consumer/producer or publish/suscriber ?
This was solid. Thank you!
Hi,
I am fresher, and new to Kafka. For storing those topics we need persistent storage and consuming a message don't delete that message. Now my question is suppose I have X ammount of persistent storage and producer produces X ammount message in Y days, what will happen to my storage after Y'th day?
Can the offset be synced across partitions so that we can have serial processing of the data ?
Very helpful. As a bizarre side note, the speakers voice sounds a lot like Weird Al Yankovich to me. Which is obviously a very good thing.
awesome video. Thanks
@10:56 It is cornflower blue. I know this because Tyler knows this.
Thanks for this great video! I have a few questions that hopefully someone could help clarify :)
1. How are brokers replicated? It sounds like it's async replicated, hence I imagine when the leader failover there would be some small amount of msg loss (because replica would always lag primary a tiny bit)?
2. When the broker receives the msg, does it write to log immediately or does it do some kind of in-memory buffering and write by small batch? And if so what happens to the non-flushed messages if that broker crashes?
Just to clarify, I'm not criticizing Kafka, it is a great tool and I really liked it while working with it in my previous job. But I'm just curious because I've heard various techtalks about how kafka is used in various products -- e.g. nu bank which is a financial startup uses kafka and from their talk I had this feeling that they rely on kafka for being 100% reliable (as in, not losing messages), which surprises me.
3. One last noob question, apparently Kafka's great capability for supporting high write throughput is partially due to its sequential write, hence avoiding random disk seek. But given that consumer almost always consumes the message with a slight delay, does it mean whenever a consumer pulls new messages it breaks this nice sequential mechanism (because we need to seek to a different disk location than the end of the log file)?
Thanks!
To answer question #1 (after watching the next video in the series), apparently one can tune the replication iin Kafka to only have producer acked after the msg is replicated to all replicas.
Hey have you found the answers to your questions #2 & #3? I don't think any system can be 100% reliable but considering industry wide adoption I believe they must be doing something really well.
I would love to know their solution to these fundamental data problems
I came for information and received quite some laughs as well.
thanks for the video, but you made a mistake about the colors:
#71cc01 - not green
#ee9f00 - not orange
#7bb1fe - not cornflower blue
Nicely explained. Thanks
Great talk!!
At 4:53 why you have `n` everywhere? That might be confusing since it is a kind of implication that we have equal number of components on all the levels.
Mustard is yellow, not orange. Great video
love the video
Wait, but how do the consumers of the consumer_offset topic keep track of where they are?
i suppose that Kafka is like those newsstands where some of us still go to to get a newspaper / magazine
Very good
wowwwwwww this video is sooo freakinggg goood
and this summarize my master degree in 24 mins.
21:24 now you confused me. I thought that the consumers used long polling, but you described a short polling mechanism.
Fantastic video @Confluent and @TimBerglund. Confluent is amazing in the way that they are making kafka easier for everyone to learn.
What s the maximum file size that Kafka can process?
I had no idea Bill Burr is so good with Apache Kafka stuff as well...
Thanks man