Видео

You can configure a MariaDB Galera cluster with MaxScale for high availability, but exact read/write traffic splits are not currently natively supported by MaxScale. However, you can prioritize nodes using the rank parameter (Node1 as primary for writes, Node2/Node3 as secondary for reads) which sounds very close to what you want, if I understood correctly. For true auto-scaling, you might want to use Kubernetes to dynamically add/remove nodes based on load (check the MariaDB Kubernetes Operator).

Hey thanks for the feedback. I'm glad you enjoyed the video!

If you are referring to installing MariaDB, you can find binaries and Docker images. If you are referring to installing additional storage engines, you have to consult the documentation for the storage engine that you want to install.

@@alejandro_mariadb alright thank you :D

Super useful when migrating from Oracle and others to MariaDB

Thanks for watching!

Thanks for watching

Thanks for your feedback. LONGTEXT is used to store JSON in MariaDB because JSON documents can be quite large, and `LONGTEXT` can handle up to 4GB of data. The `JSON` data type is a synonym of `LONGTEXT` btw. To validate JSON, the function `JSON_VALID()` is used. It checks if a string is a valid JSON document, for example: `SELECT JSON_VALID('{"key": "value"}')` returns 1 if valid. You can use this function, for example, in a constraint check.

@@alejandro_mariadb Thanks for the reply, I tried to add a constraint with and without "ADD" (seen "working" examples both ways), and with and without naming it, "PhpMyAdmin" refuses to add it , one attempt: ALTER TABLE FinderPlant ADD CONSTRAINT xxxx CHECKED( JSON_VALID(PhotosJson) );

​@@DennisBareis The correct syntax is `CHECK` not `CHECKED`: ALTER TABLE FinderPlant ADD CONSTRAINT xxxx CHECK (JSON_VALID(PhotosJson));

@@alejandro_mariadb Thanks but even your syntax fails, I have however successfully added it 3 times as I found a way to view them: SELECT * FROM information_schema.check_constraints WHERE table_name = 'FinderPlant';

@@DennisBareis I think it should be "CHECK" not "CHECKED".

Same here with MacOS

Thanks for watching! It is a cool t-shirt right? And it's pretty cool that you noticed it!

Awesome! It can be really useful in certain situations.

I'm glad you liked it!

Replication is such an interesting topic and a strong feature of MariaDB. I'm glad to hear that the video was useful!

Did you provision ColumnStore? docker exec -it mcs1 provision mcs1

@@alejandro_mariadb yep. I found out that I have to manually start cluster after every docker container restart.

@@palychbasist I wonder if there's an option to start it automatically...

That was done in post production. Just a basic zoom-in animation to the corner.

For database companies, commandline utilities and terminals are kind of the same experience across operating systems, so it's a good way to show things. When you pick one IDE then people that use a different IDE and feel like the video isn't for them. Also I think you have a decent size community of developers that just don't use IDEs at all, and work at the command line, or with text editors, and do a lot of bash scripting

SAP HANA is an interesting product that stores data in memory. This likely requires more expensive hardware than with MariaDB. In general, data storage is less cost-efective with in-memory solutions. Memory requirements scale with your data. There could also be advantages in ColumnStore over SAP HANA regarding backups, management, and even security-topics worth exploring during evaluation phases. On the other hand, yes, with CS data is copied so that it can be stored in an optimal way for both transactional and analytical workloads.

the columstore docker image does run on arm CPUs hub.docker.com/r/mariadb/columnstore

My company was evaluating it for use in production and although it seems fast when you do something it doesn’t like the entire db simply crashes. Would love to talk someone about it

​@@tbone587 Any useful information in the logs? Please join the MariaDB Community Slack where you'll get more eyes on your question.

She's speaking English in the video - perhaps try enable subtitles? support.google.com/youtube/answer/100078?hl=en&co=GENIE.Platform%3DiOS

I like HAProxy. It's pretty cool. However, MaxScale has many features not present in HAProxy. Just to mention some of them: Automated failover for MariaDB Replication, query-based load balancing with ReadWriteSplit Router, connection-based load balancing with ReadConnRoute Router, data Import from Kafka and export to Kafka, cache filter for improved SELECT performance, security and traffic controls with auditing and SQL injection protection, web-based GUI for configuration (with REST service available too)

We are switching to MaxScale because HAProxy can't do things like MaxScale does. HAProxy is basic and used mainly for HTTP. A database proxy is better with a database. HAProxy is better for HTTP.

A database proxy (like MaxScale) is still extremely useful even on IPv6. I can add features such as HA with in-flight query migration, CDC and data ingestion to and from Kafka, transparent data masking, and more.

@alejandro_mariadb that you can add a proxy to an ip v6 network, yeah, but it's not required. You can resolve the same with Microsoft power automate it is faster working in every region, and suport multiple applications 😉

@@adriancoanda9227 That sounds super cool. Here I'm talking specifically about "database proxies". Check my "Database Proxies Crash Course" on this channel if you want to learn more about this.

Thank you for watching!

I agree. There's a Jira issue for traking this development (MDEV-8100). Please, feel free to vote and follow it. Meanwhile, you can somewhat mimic this functionality using either temporary tables, or the JSON_TABLE + JSON_OBJECTAGG functions to pass tabular data between a function and its client (caller). It's not the most elegant way, but it gets the job done.

@@alejandro_mariadb really interesting solution, thanks. Also checking the Jira issue

Not that I'm aware of. I failed to understand a scenario where this is needed but the closest I can think of is with the BLACKHOLE storage engine. This storage engine doesn't write data at all. However, binary log events are recorded. So, if you have a primary server with a table using, say, InnoDB but BLACKHOLE in the replicas, then the writes only happen in the primary and not in the replicas. Conversely, if the table uses BLACKHOLE in the primary and InnoDB in the replicas, when you run INSERT statements on the primary, the rows are inserted only in the replicas. I'm aware that this is not exactly what you are asking, but wanted to share this ideas.

Is this for a CQRS type architecture? Where you have a database that is handling writes only so they're faster and dedicated then replicating or something? Curious if you bounce back to this question if you can give more detail. I've got a few thoughts on how to do it if that is what you're aiming for.

MaxScale can dedicate one node “Primary” for write only and other nodes for read only, but if no “read only” nodes remain in the cluster, MaxScale will be forced to use the remaining node for read and write.

There's an "Stop Execute" button that looks like a red on/off icon.