ArchitectureA supervised JVM-class runtime — OLTP on seven engines, OLAP on three. AI-native, MCP-native, observable as plain SQL.Read the architecture
Está viendo la edición Perú. Está viendo la edición Colombia. You're viewing the Pakistan edition. Cambiar a la edición global →Cambiar a la edición global →Switch to the global edition →
ObservabilityProduct Release2 min read

Running work as records: sessions, cursors and scripts you can read and cancel

The observability catalog gains session and scripts tiers. A cursor shows what it asked, a running script shows its code and line, a delete on the row cancels it, and an hour of metrics is kept every five seconds.

The observability catalog described the JVM, the connection pools and the caches well, and the people and programs using them poorly. Sessions were grouped by the server that stored them, a running script was a thread name, and a cursor's query had to be decoded from a URL. This release adds two tiers that treat running work as records.

Session tier

  • A schema of their own Sessions, the session cache and cursors stand apart from the web server storing them. The per-node login cache stays where it is.
  • Restored sessions counted Sessions restored from the data store after a restart are adopted by the tracker, so an administrator sees them and can force a logout.
  • A cursor says what it asked Five new columns: the query parameters as JSON, the effective where clause as SQL, the title, the age and the error. Each opens in its own editor.
  • Scripts launched by a session A table lists the scripts an HTTP session started, with the user and the admin flag.

Scripts tier

  • JavaScript and Python alike The scripts schema names the language of each run: JavaScript, Python, Groovy or 4GL.
  • Code and position A running script shows its whole body, the first ten lines as a head, and the source and line it is executing, from a statement-level listener that collects no values.
  • Cancel from the row Running scripts and cursors open as records. Deleting the row cancels the run or closes the cursor, and deleting a session row is a forced logout.

History and errors

  • An hour of headline metrics Eight series, heap, CPU, threads, exceptions, SQL rate, GC time, HTTP requests and HTTP 5xx, sampled every five seconds for 720 samples. Every other rate in the catalog was a trailing average of the moment it was read.
  • Errors with their trace Per exception signature the catalog keeps the message, the last log id and a bounded stack trace, keyed by server and database, so the same SQL error on two databases is two lines.
  • Web tier counters Request totals, response classes, request time, and the thread pool's own low-on-threads verdict.

The catalog pages are regenerated from the same source. The cache catalog is drawn as a treemap nested by server and database, the JDBC tier as a topology that folds to counters above sixty pools, a Memory page adds off-heap buffer pools and allocator segments, and every list opens code in the editor of the row's language.

On the platformOperationsPlatform

All posts