MariaDB Foundation Newsletter – 15 October 2026

The Ecosystem Hub passes 200 entries

The MariaDB ecosystem has always been larger than any one company or product. Until this year, that was hard to see. The Ecosystem Hub, launched in March as a curated, open directory of products, services and open source projects that work with MariaDB Server, has now passed 200 entries. They span cloud and hosting, monitoring and observability, backup and recovery, security, migration, connectivity, DevOps, analytics, AI, application development, CMS, e-commerce and complete business applications.

The next step is to connect individual technologies into useful architectures. The first Solution Stack, the Privacy-First Collaboration Stack, combines MariaDB Server, Nextcloud and Passbolt. Stacks for sovereign infrastructure, e-commerce, multicloud independence, high availability and open web performance are in development.

Listings are free. If you build or maintain something that runs on MariaDB, it belongs in the Hub, and the 200 entries already there are a good argument that people will find it.

Open Source milestone: Ecosystem Hub hits 200 solutions (8 Oct) →

Sponsor roundup: Frappe and Solanica join as Silver Sponsors

Sponsorship gives MariaDB Foundation the resources to support MariaDB Server’s openness, adoption and independence. Three Silver Sponsors have joined since the start of October: Elestio, which we welcomed in our last issue, and now Frappe and Solanica.

Thank you to all three: your support keeps MariaDB Server independent and community-led.

MariaDB Vector, in Laravel and in production

Laravel 13 supports MariaDB Vector natively

If your Laravel application runs on MariaDB, you can add semantic search without a second database, an external vector service or an extra package. Since Laravel 13.28, vector columns, vector indexes and similarity queries work on MariaDB 11.7 or later with the mariadb driver. Until this summer, Laravel’s built-in vector search assumed PostgreSQL and pgvector.

A large part of that work came from the community. Mohamed Touhami Rhaima, a PHP developer in Tunisia, made vector queries compile to VEC_DISTANCE_COSINE and got Laravel’s CI running on MariaDB 11.8 LTS and 12.3 LTS, where before only MariaDB 10 ran the database suite and the vector tests were silently skipped. Contributors michielvaneerd, eas4ai and xurshudyan added vector indexes, the AsVector cast and dropVectorIndex(); Taylor Otwell reviewed and merged all of it. “I keep going because every gap closed helps other Laravel developers who run MariaDB,” says Mohamed.

Laravel 13 now supports MariaDB Vector natively (7 Oct) →

Bagisto: semantic search, tuned and measured

The team behind Bagisto, the open source Laravel e-commerce platform, tested semantic product search on MariaDB Vector inside their existing database, on MariaDB 11.8 LTS in a container limited to 4 CPUs and 2 GB of RAM. They did not stop at showing it works: they measured indexing, latency, concurrency, filtering and recall, and the effect of tuning the vector cache, the InnoDB buffer pool and the HNSW index. The headline finding is the index parameter: raising M from the default 6 to 16 lifted recall from 59.2% to 94.5% at the same latency. Keyword search, as the article puts it, “fails the moment a shopper stops thinking like a database.”

Bagisto MariaDB Vector Semantic Search (13 Oct) →

Performance testing in the open

Performance engineering inside a database project is usually private. The Foundation’s Test Automation Framework (TAF) is an attempt to change that, and October brought three steps.

TAF now speaks PostgreSQL

TAF was designed so that database makers could plug their own engine into it. After MariaDB and MySQL, a full PostgreSQL plugin has landed, funded by sponsor Virtuozzo and written by Lukas Oliva as a self-contained Perl module. HammerDB TPROC-C and TPROC-H run natively; sysbench runs once compiled against the target PostgreSQL. The invitation stands for any other database: build a plugin, get automated installs, structured workloads and concurrency sweeps.

TAF adds PostgreSQL (5 Oct) →

A regression reproduced, and a fix confirmed

Mark Callaghan filed MDEV-37992, a regression that appears only under specific LRU and buffer pool conditions and needs a memory-resident workload to expose. Using TAF, Jonathan Miller reproduced it on 11.8, 12.3 and 13.1 with a 10 million row, 40 thread, 90 GB buffer pool workload, then ran the patched commits from MDEV-33966. Throughput came back on every branch. The properties file and driver script are in the TAF repository, so anyone can run the same test.

That is the point of the new Performance MDEV Test Case Library: make performance change detection public, reproducible and teachable. The first case, MDEV-32750, packages an Adaptive Hash Index behaviour change found during a 10.4 to 10.11 migration. As Jonathan puts it, “A drop might be a real problem. A gain might be a real problem. A surprise is always worth investigating.”

Reproducing and Validating MDEV-37992 with TAF (14 Oct) → The Performance MDEV Test Case Library (8 Oct) →

173 security reports in two quarters

Some recent releases arrived later than expected, and Frédéric Descamps explains why. For years MariaDB received a handful of security reports a quarter. Then came 81 in one quarter and 92 in the next. These were not AI slop: almost all were good reports of real issues, many in old code that AI-assisted tooling helped researchers re-examine, and all were fixed. Each one means reproduction, review, backporting across maintained branches, an advisory and a CVE request. The project now publishes its own advisories on GitHub rather than waiting weeks for CVE assignment. More reports mean more research and better tools, not weaker software. Update to the latest maintenance releases.

When AI Finds the Bugs We Missed (2 Oct) →

Also worth a look

Sergei Petrunia on how MariaDB 13.0’s optimizer hints go beyond MySQL’s (10 Oct): path-based query block addressing works through views and derived tables, with syntax compatible with TiDB. On the main branch, Arcadiy Ivanov’s 24-commit batch bringing BLOB, TEXT, JSON and GEOMETRY support to the MEMORY engine (MDEV-38975 and follow-ups) was pushed on 11 October, the largest external contribution of the fortnight; Roman Nozdrin landed replication fixes for the DuckDB engine (MDEV-40957); and a new --validate-config option checks a configuration without starting the server (MDEV-31527, commit by Abdelrahman Hedia).

Connectors fixed, server releases ahead

No MariaDB Server release shipped in the first half of October. The current line-up is 13.0.2 Stable, 13.1.1 Release Candidate and the 13.2.0 Preview, with the Q4 maintenance releases for the LTS series still to come.

Connector/ODBC and Connector/C++

Connector/ODBC 3.2.10 and 3.1.24 and Connector/C++ 1.1.9 and 1.0.8 shipped on 5 October, all Stable. These are security releases: they close several memory-safety issues where a malicious server could make the driver write outside its buffers, and SQL injection paths through crafted column and savepoint names. One behaviour change to note in Connector/C++: multi-statements are no longer allowed by default, and the new allowMultiQueries option turns them back on.

Updated MariaDB C++ and ODBC Connectors (5 Oct) →

From MariaDB plc

Bengaluru, and beyond

Open Source India

The Foundation was at Open Source India in Bengaluru on 7 and 8 October, represented by Simona Aleksandrova and Roman Nozdrin. Talks covered the database as a platform, the path to MariaDB extensibility and the DuckDB storage engine, and a workshop walked attendees through building MariaDB plugins. At the booth, Slava Rozhnev’s MariaDB quiz did what it did in Amsterdam: it turned a quick visit into a conversation. Special thanks to Anway Durge and KC Jagadeep of Ceymox, who supported the booth and took on a good share of the technical questions.

Bengaluru was a start, not a stop. The Foundation is building a series of events around India this winter, beginning with our first India meetup, which will be held online. Stay tuned.

Community Spotlight: Slava Rozhnev and SQLtest.online

Open source runs on people, and not only the ones who commit code. Slava Rozhnev has been building SQL quizzes with the Foundation for a while, and his platform, SQLtest.online, lets anyone learn SQL by writing real queries against real databases, with 460 tasks, six interface languages and no account needed. He runs it independently, and unlike a booth it does not pack up when the event ends.

Community Spotlight: Slava Rozhnev (12 Oct) →

JOIN TABLES: Simon Mudd

The second episode of JOIN TABLES is out: Simon Mudd, a familiar name to anyone who has run MySQL or MariaDB at scale, in conversation with Kaj Arnö, Executive Chairman of the MariaDB Foundation.

JOIN TABLES – foundation JOIN simon_mudd (14 Oct, 13 min) →

500 GB from RDS to EC2, with a way back

Pradip Shah of luroConnect documented moving a 500 GB production Magento database from Amazon RDS to self-managed MariaDB on EC2, with two conditions from the client: near-zero downtime, and a near-zero-downtime rollback path. The cutover through ProxySQL took under ten minutes. Monthly cost fell by 83% year on year, and, as the author is careful to say, cost was the sole driver; faster uncached pages were the outcome. It is not an argument that every workload should leave RDS. It is a well-documented account of how to do it when the numbers say so.

Migrating a 500 GB Production MariaDB Database from RDS to EC2 (5 Oct) →

Oracle MySQL

Percona

VillageSQL

Elsewhere

Influence is not the same thing as personal ownership of the project.Anna Widenius in ODBMS Industry Watch, 1 Oct

Roberto Zicari interviews Anna Widenius and Kaj Arnö on governance, open source and the future of MariaDB: why open source code without inspectable governance is only half open, and why, in Kaj’s words, “PRs languishing in the queue are probably the best early warning to look out for.” Influence Is Not Ownership

Full listing at mariadb.org/events