01 · INGEST
AWR parser
12-step pipeline. 173 sections, 123 tables, 4,600 rows extracted per Oracle 19c RAC AWR. Detects parser regressions before they reach you.
Single-instance, RAC, Exadata, CDB/PDB, multi-PDB. All parsed with the same pipeline.
We don't ship dashboards — we ship the work. Every tile below is wired to the same parser, same classification catalog, same workspace.
12-step pipeline. 173 sections, 123 tables, 4,600 rows extracted per Oracle 19c RAC AWR. Detects parser regressions before they reach you.
Servers and databases derived on the fly from nine signals per AWR — DBID, DB_NAME, hostname, cluster_name, instance #, container, version, platform, machine specs. No CMDB. No setup.
One row per sql_id across your entire fleet. Sort by elapsed time, executions, buffer gets, I/O. Filter by category. Drill into any sql_id for full text, plan drift, per-AWR breakdown.
Platform regex catalog tags Oracle internals, vendor apps (EBS, PeopleSoft, Siebel), ORMs (Hibernate, Django, ActiveRecord), and monitoring noise (OEM, Datadog, Statspack). Workspace overrides win for your custom SQL. Conflict detection prevents hiding real bugs.
Workload, memory, and efficiency over time per database. AAS, SGA/PGA stacked, buffer hit %, soft parse % — all charted across every AWR you've uploaded.
Cross-database analysis on shared hosts. SGA stack (sticky), PGA per-DB lines (volatile, honest), AAS heatmap, roster gantt colored by sample density. Built for mixed cadences across databases.
Capture STS from your DB, ship structured. We compute static SQL Performance Analysis: regression / improvement / unchanged classification across two captures. Plan drift on every sql_id.
Servers and databases aren't configured — they're derived from the AWRs you upload. Every report contributes nine signals, and the inventory rebuilds itself on every upload.
Drop an AWR HTML into a workspace. Twelve steps later you have a normalized, queryable representation — sections, tables, fingerprints, version, architecture, corruption flags. Nothing more, nothing less.
The parser produces facts only. No verdicts baked in at parse time — verdicts come from the UI layer, where you can see exactly which facts produced them. 96 supported version/architecture combinations, with regression tests on every step.
Shareable. Bookmarkable. Linkable from your runbooks. One row per sql_id across your fleet, with elapsed/executions/buffer-gets/I-O, category pill, and a drill-in to full text, plan drift, and a per-AWR breakdown.
Classification runs through a platform regex catalog with workspace-scoped overrides. Tenant admins can hide their custom SQL from "noise" categories while keeping the platform catalog up-to-date — conflict detection prevents you from accidentally hiding a real bug.
We solve that. Sparse cells mean "no AWR in this window," not "we guessed for you." SGA is sticky and stacked; PGA is volatile and honest. The roster shows you what you actually have, before the math runs.
Oracle AWR is the most important diagnostic output in the database world, and reading one is a chore. Reading fifty is a job. We built GatorRed because every Oracle DBA we know was still opening AWRs one HTML at a time, Ctrl-F'ing for the wait event they remember, and pasting screenshots into tickets.
Our pipeline is open-source-grade rigorous: 96 supported version combinations, RE2-based SQL classification, asymmetric workspace overrides with conflict detection, soft-deleted file handling, and a parser that produces facts only — never verdicts. Verdicts come from the UI layer, where you can see exactly what produced them.