Apache Cassandra Apache Cassandra World Party | April 2021, 28 Apr 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Cassandra Robustness: Errors I made and you can't anymore! | Apache Cassandra World Party 2021

Description

Presentation from the first Cassandra World Party on April 28, 2021. Learn more about Apache Cassandra at https://cassandra.apache.org/.

Carlos Rolo of Instaclustr presents “Cassandra Robustness: Errors I made and you can't anymore!” on errors he made through his 10+ years with Cassandra, across different versions, and their impacts. Better yet, he'll dive into how robust Cassandra is right now!

About the Presenter
Carlos is a technical account manager at Instaclustr and has been working with Cassandra for 10+ years. He has a passion for open source and distributed systems.

Apache Cassandra is an open source NoSQL distributed database trusted by thousands of companies for scalability and high availability without compromising performance. Linear scalability and proven fault-tolerance on commodity hardware or cloud infrastructure make it the perfect platform for mission-critical data. Get started: https://cassandra.apache.org/quickstart/

The Apache Cassandra World Party 4.0 was a virtual, one-day celebration on Wednesday, April 28, 2021 in preparation for the launch of 4.0, the most stable database release ever shipped.

Join the Cassandra community: https://cassandra.apache.org/community/

#ApacheCassandra #database #NoSQL

A

This is uh carlos rolo. uh I've known carlos for a really long time and he'll always tell cringy stories about how we met, but um carlos has been known as the uh the architect or the uh the the, not the architect, the archaeologist there. He is the archaeologist.

A

We give him such a hard time because he, when he worked at pythian his job, was to find the oldest cassandra clusters and get them back in order.

B

Welcome carlos my name is carlos lovo, I'm a technical account manager at easter, foster. I've been with cassandra since.

B

0.8 in production, my former colleague renee anthony's, pulled me into the community and patrick helped me a ton uh while I'm here and like my my guiding lives in the community since since I joined so so. The first thing I you don't have to fear anymore is counters. uh Counters are no longer a bad. It's such a bad situation uh in early one version, 0.8 0.9, you restart cassandra.

B

Without clearing the commit log, your counters will be recounted and that's a bug that was 16 1.1.3 and there was no really way to roll that back because you don't really know what happened.

A

B

uh I upgraded cassandra, restarted the notes and guess what I didn't know what the hell happened with my data. So it's not a fine way to discover encounters with a bad thing and but you don't like it again, you don't have to worry it's done and it's uh it's fixed, so good for a while.

B

Okay streams stream was, um you know, stream is something that should happen. Naturally, early 2.1 releases had a weird bug that streams would time out. So it's not a fine thing. If you were like adding a new data center, both striking bootstrapping a couple of new nodes, you have to like be really attend.

B

What's going on on network because you just might have to restart, you just have to restart the note yeah that one was nasty like oh, it's crash three star, okay with realized views. I have a personal grudge against this one. It's the only full down time I had with cassandra. They are disabled. By now we should keep it disabled.

B

My colleague john schulz was on that night where everything had to go down and again now, if you're protected, unless you really want to mess with it avoid it. Please.

B

And the wrong partitioner yeah there was a couple of other partitions. Random was the first one. It was one called byte. Ortho partitioner! Don't worry about that one anymore! It was really good. If you use time based data because then your log, your data is always hot. Some node was always hot. It was amazing how terrible that thing was so no longer. Is there not enable it? Don't get back, don't think about it.

B

Stick to the default features.

B

Solid, no two chuckles patrick, is laughing yeah. I did it once. I hope nobody else did it.

B

And uh yeah the basically you had no v nodes. You wanted v nodes, you should do this shuffle and it will break your cluster into a weird state. So yeah, don't don't do it stable cassandra. 4 is going to be a great release. It's stable! It's all about practice and stability, so you don't have to go through a presentation like this there's a lot of testing it's getting released with a green board and to be fire is the first time I'm going to write a blog post and move to a top zero release.

B

I did it never did it before. So it's always like go 2.6.

B

Now you can go 2.0 thanks, juan thanks everyone and ask your questions on chat, share your other stories on early cassandra releases and if you have a really old quest that you need to sort it out, you can always fingers I'll try to help you or maybe not.

A

Oh god yeah that was, uh I yeah I was cracking up just because I think I remember all those stories, no tool shuffle yeah great idea. The jira looked great yeah carlos. Thank you very much. We're gonna bring nick on.