Visit Azul.com Support

Ic On Premises 1.1.2

Table of Contents
Need help?
Schedule a consultation with an Azul performance expert.
Contact Us

Azul Intelligence On-Premises 1.1.2

Release date: November 17, 2026

Azul Intelligence On-Premises 1.1.2 moves the observation dataset to a newer embedded DuckDB engine, which speeds up inventory reads, wave queries and ingestion. It also fixes server errors on machine pages and updates bundled security libraries.

You can upgrade in place from 1.0.0, over the same data directories, as described on Upgrading Your Installation. No database migration runs, no separate conversion of the DuckDB file is needed, and you don’t need to change any setting. However, one default changes, for the query engine’s worker threads: see Behavior Changes below. .

Important
Once 1.1.2 has opened your data directory, running 1.0.0 on it is unsupported. Take a complete backup before the upgrade and keep it until you accept the upgrade, because going back to 1.0.0 means restoring that backup.

What’s New

  • Third-party notices: The bundle contains THIRD-PARTY-NOTICES.txt, which lists the third-party software that the bundle includes, with its licenses. These components are also listed on Third Party Licenses.

  • Where to get vulnerability data: While no vulnerability data is loaded, Settings > Vulnerability Data and, for an Admin, the Security Exposure page link to the documentation on where to get it. See Where to Get the Dataset.

Behavior Changes

  • Fewer query engine threads by default: With EVENTS_DB_READ_THREADS unset, the embedded DuckDB engine runs on half the CPUs that Java sees, rounded down and at least one, instead of all of them. Azul measured reads on hosts with 8, 16 and 32 vCPUs, and half the threads were never slower than all of them. When you set EVENTS_DB_READ_THREADS, the engine uses your value as given. The startup log line DuckDB read tuning in effect: … threads=… shows the value in effect.

Improvements

  • Faster reads and ingestion with DuckDB: Inventory pages, machine details, runs, Forwarder machine counts, wave queries and ingestion are faster. A repeated inventory page or Forwarder count comes from its retained result, which keeps throughput up under load.

  • Clearer labels: The Inventory search box says "Search host…", which is what it filters by. The Matching installations card reads "Installation count, percentage of fleet, and license distribution."

Bug Fixes

  • Installations without a path: A machine’s page, in the web interface and the public API, opens when one of its installations has no known path, instead of returning a server error. The page lists that installation without a path and with no exclusion marker.

  • Invalid record IDs: A record ID that is not a UUID returns "not found" instead of a server error, in the web interface and the public API.

  • The inventory list and the machine detail agree on the machine’s latest known fields, including when its newest observation omits host or container information.

  • The machine detail keeps all Agent and scanner Forwarders while the inventory is rebuilding.

  • Wave holder values and row ordering stay stable when observations tie.

Security

  • Updated Quarkus to 3.40.1.2, which brings jackson-databind 2.21.8 and brotli4j 1.24.0 with their security fixes.