Konstantine Karantasis

66 posts

Konstantine Karantasis banner
Konstantine Karantasis

Konstantine Karantasis

@karantasis

Distributed systems, s/w craftsmanship, concurrency & parallelism. Maybe cinema, coffee & cycling too. Any resemblance to real opinions, is purely coincidental

San Francisco, CA Katılım Ocak 2014
179 Takip Edilen357 Takipçiler
Konstantine Karantasis
Konstantine Karantasis@karantasis·
"More importantly though, data streaming is about connectivity", well-written and insightful thread overall. Highlighting this quote because I'm a bit biased too 😀
Jay Kreps@jaykreps

1. Thoughtworks notes that "Kafka continues toward its status as a de facto standard." noting that Kubernetes, Kafka, and the CSPs are becoming stable layers in the next gen stack and churn around alternative platforms seems to have waned. thoughtworks.com/content/dam/th…

English
0
0
5
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
Definitely a highlight of this summer for me and a very rewarding experience to work with the Apache Kafka community to release version 3.0.0
Apache Kafka@apachekafka

Apache Kafka 3.0 is released! 🎉 Another step closer to remove ZooKeeper dependency. And of course many other improvements, bug fixes, and new features. Thanks to 141 contributors and our RM @karantasis! For more details check out the release blog post: blogs.apache.org/kafka/

English
1
0
6
0
Konstantine Karantasis retweetledi
Jay Kreps
Jay Kreps@jaykreps·
1. A quick reflection on Confluent's IPO today and the journey so far (a thread!).
English
25
191
1K
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@DanielOsvath wrote about his recent experience with Kerberos and this tutorial can save some of us a lot time during troubleshooting. Good security is an integral part of every decent data integration solution and lots of Kafka connectors support Kerberos authentication already.
Confluent@confluentinc

Kerberos authentication is widely used in today’s client/server applications; but getting started without experience can be a challenge! @DanielOsvath provides a simple setup for a Kerberos test environment with SSH in today’s blog. Check it out: cnfl.io/containerized-…

English
0
0
1
0
Chris
Chris@criccomini·
@karantasis Super confusing since docs say default is "null", but the comments on the docs mention SimpleHeaderConverter.
GIF
English
1
0
0
0
Chris
Chris@criccomini·
Ugh. APM is inserting headers into Kafka records. Header causes Kafka connect task to halt with 100% CPU one in every N billion messages. Kafka console consumer doesn't have the issue. Three times in three weeks (including during son's birthday party). How's your week going?
English
10
4
38
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@criccomini I totally agree with you on that. It's a general default but with surprising side-effects. I've suggested documentation notes/additions at a minimum and we should reconsider it as a default, indeed.
English
1
0
0
0
Chris
Chris@criccomini·
@karantasis Also, not convinced SimpleHeaderConverter is a good default. Too much magic. Too slow.
English
1
0
1
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@criccomini The converter might be crashing or spending too much time on non-UTF encodings. I've noticed abnormalities with other (less popular) encodings in strings.
English
1
0
0
0
Chris
Chris@criccomini·
@karantasis Yea, we don't even care about these headers. I also have no explanation about why only certain records are causing problems.
English
1
0
0
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@criccomini My benchmarks have revealed that ByteArrayConverter is 4x-5x faster and StringConverter 2x-3x (with lower CPU), so if you can use these 👍. But if none of these converters fits your needs maybe a custom implementation (this class is configurable) could offer peace of mind.
English
2
1
3
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@criccomini SimpleHeaderConverter is meant to be general, but as its name reveals it's also simple. In high volume it's general approach of inferring types comes with a higher CPU cost.
English
1
1
1
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@apachekafka I'm thrilled. Thank you all! It's an honor and a great responsibility. I'm really looking forward to serving the Apache Kafka community and its innovative technology on this new role!
English
0
0
8
0
Apache Kafka
Apache Kafka@apachekafka·
We are happy to announce that @karantasis accepted our invitation to become an Apache Kafka committer. Join us to congratulate him!
English
9
13
106
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
@gwenshap Lol. And while I was thinking... nothing can stress me out before the talk at this point 😂
English
0
0
2
0
Stéphane Derosiaux
Stéphane Derosiaux@sderosiaux·
With the latest improvements in Kafka Connect (Incremental Cooperative Rebalancing), does it make sense now for a large company to provide a large Connect cluster for all its internal projects, to avoid lots of Connect clusters? (maintenance, scalability, support, run etc.)
English
4
3
10
0
Konstantine Karantasis
Konstantine Karantasis@karantasis·
Parquet support for @confluentinc S3 sink connector got merged. Highly anticipated, it was accomplished in one of my favorite ways: as a top quality community contribution! Thanks Tony Wei and Ryan Eakman! Available from source on github now, will ship with Confluent Platform 5.4
English
2
15
31
0

Keşfet

@asc89 @tlberglund @apachekafka @DanielOsvath @criccomini @elonmusk @BarackObama @taylorswift13