AGGREGATE Beta v2026.11.03: Scalability and alerts

This week, Beta v2026.11.03.

Release notes: https://github.com/Subschema-LLC/aggregate/releases/tag/v2026.11.03

This release adds alerts and multi-server operation. Both came from pilot feedback. Pilot organizations have run AGGREGATE on real websites since October, and two areas came back from that use: notification when data stops arriving, and operation beyond a single server.

Alerts

A removed tag or a broken consent banner fails silently. The site looks fine and the numbers are absent.

A new Alerts section is available to administrators. A rule states a condition: a website records no events, or fewer than a set number of events, in the last hour, day, or week. The first condition covers a tag that stops sending. The second covers a consent banner fault, where events continue but the enhanced events stop. Rules also cover tracking failures.

Alerts are delivered by email and by signed webhook. Each alert has a firing notice and a recovery notice, and an administrator can send a test alert first. The payload contains no visitor data. Alerts are off by default, behind a feature flag. Rules are set on the page, in YAML, or from the command line.

Scalability: server roles

AGGREGATE now runs on several servers that share one database. The administration server runs the dashboard, setup, updates, and the scheduled task. Collection servers run the collection endpoint only, behind a load balancer. Adding a collection server adds collection capacity. Worker servers are optional and run queued processing.

Collection servers pull their configuration from the administration server. The load balancer configuration is printed by a single command. The other servers update files-only, which keeps releases compatible across the servers. Setup guides are included for Linux servers and for Plesk.

The collection endpoint also changed in this release. It builds only the parts collection uses. Event rows are written with a single database insert instead of the ORM. Rate limit buckets are kept in APCu, measured below the overhead limit set for them.

The load testing notes in the repository record how capacity changes as collection servers are added, and where the shared database becomes the limit. The multi-server guide has been followed in continuous integration, on one machine. It has not yet been run on separate servers.

Smaller items in this release: sign-in cards use the navigation color, queued messages older than seven days are deleted, and an update requests a PHP reload and reports when the reload did not happen.

Upgrading

A single server updates in the usual way. In a multi-server setup, the administration server updates first, then the other servers files-only, with collection servers updated one at a time. All servers run the same release.

Still beta, still installed from the release ZIP or Docker, no independent security or privacy audit claimed. Download from the AGGREGATE page: https://subschema.co/products/aggregate-analytics-by-subschema

AI transparency: originally drafted by AI and edited/revised by the poster

Back to Blog