DrWatson: root cause analysis for Informix and enterprise systems
Beyond Detection. Root Cause Intelligence.
DrWatson is Oninit’s proprietary diagnostic engine. It uses symbolic reasoning, expert diagnostic knowledge and evidence correlation to find the causes of complex Informix and enterprise system problems, not just report their symptoms.
It is built on decades of Oninit Informix expertise. Every finding links back to the evidence and rules behind it, and DrWatson’s core analysis does not rely on a third-party generative AI model.
See How DrWatson Works Request a Demonstration
Root cause analysis, not just error detection
Monitoring tools and log analysers can tell you that something is wrong: an assertion failure, a hung thread, a spike in waits. They rarely tell you why. Real incidents leave evidence scattered across the database engine, the application server, the JVM and the operating system, and the root cause is seldom in the first error you see.
DrWatson reads that evidence together. It normalises the diagnostic data you submit, correlates observations across components, and applies a structured, growing library of expert-defined inference rules to derive root-cause findings from the correlated evidence. Every finding comes with prioritised recommendations you can act on.
Slow Informix database? Find the layer that’s really at fault
“The database is running slow” is where many investigations start, and it’s rarely where they end. The cause could be the database engine. It could just as easily be the application sending the queries, the virtual machine the database runs on, or the physical host and storage under that VM. Each layer keeps its own logs, and each team usually looks only at its own.
Give DrWatson the logs from each layer and it analyses the whole stack together: application, database, operating system, virtual machine and the hardware underneath. Instead of every team proving its own layer is innocent, you get one finding that shows where the slowdown actually starts.
How DrWatson reasons
- Collect. You upload the diagnostic data your systems already produce through the portal. No agent is installed and no schema changes are made anywhere in your estate.
- Normalise. Each input becomes structured facts, using whatever level of detail the data contains. Partial data still gives useful results.
- Correlate. DrWatson relates facts across sources, so a symptom in one layer can be traced to its cause in another.
- Infer. Expert knowledge, encoded as symbolic rules, is applied to the correlated facts to derive root-cause findings. DrWatson also cross-checks its conclusions against the rest of the evidence: for example, it separates the thread that failed from the code that exhausted the heap, and won’t dismiss a quiet period as routine when exception dumps show otherwise.
- Explain. You get a report of findings, recommendations and the evidence trail behind each conclusion.
Explainable diagnostics: every finding has an evidence trail
A DrWatson finding isn’t a black box. Each finding in the report carries the ID of the rule that produced it, the evidence it rests on, and the recommendations that follow from it.
You can see why DrWatson reached a conclusion, check it against your own knowledge of the system, and keep the report for change control, audit or post-incident review. Terms such as AF, onstat and ONCONFIG are explained in the glossary.
DrWatson’s results are also repeatable. Running the same evidence through the same rule version gives the same findings, recommendations and evidence trail, because the core analysis is rule-based with no language model in the loop. If a conclusion changes, you can trace it to a change in the evidence or the rules.
Curious how the rules reach a conclusion? See How Prolog Inference Diagnoses a Database.
Proprietary reasoning. No generative AI in the analysis.
DrWatson’s root cause analysis is performed by Oninit’s own symbolic reasoning engine and diagnostic knowledge base. It does not send your diagnostic data to a large language model to interpret.
Two optional features sit on top of that analysis and use a third-party language model: a Q&A panel, for paying customers, for exploring a delivered report in plain language, and a rewording of the narrative text in WebSphere reports, which is applied automatically when DrWatson detects WebSphere content in your upload. They are separate layers. The Q&A panel works from the text of that report’s findings, and the narrative step sees only the derived narrative text, never the underlying fact base, your uploaded files or anyone else’s data. The root cause analysis is the same with or without them.
Where your data goes. DrWatson is a hosted service, operated by Oninit. Nothing is installed in your environment. You upload diagnostic files through the portal, and DrWatson never connects to or runs on your database. Each submission is stored in its own isolated area. By default your uploaded files are deleted after 30 days and the whole submission, including the report, after 90 days; different terms can be agreed. See Security & Privacy for details, and Obfuscation for what you can scrub before uploading.
What DrWatson analyses: supported diagnostic inputs
Whether you are doing Informix troubleshooting, Informix assertion failure analysis, onstat -a analysis or performance tuning, or reading a WebSphere javacore or thread dump, DrWatson does the log analysis and gives you a root cause analysis (RCA) report.
Informix
- ifxcollect data: configuration, environment and runtime metrics across collection profiles, used for system state, misconfigurations and workload characteristics. See Configuration Audit for ONCONFIG.
- AF (Assertion Failure) files: engine failures and internal errors, matched against known failure patterns to find likely root causes. See Health & Errors.
- onstat -a output: memory, locks, sessions and I/O, used to find bottlenecks, contention and resource constraints. See Performance & Tuning and the onstat reference.
- Query plans (
sqexplain.out): output from SET EXPLAIN ON, used for query plan analysis.
Beyond Informix
- IBM WebSphere logs, JVM thread dumps, heap dumps and javacores: hangs, deadlocks, excessive GC, memory leaks, thread contention and configuration issues.
- Operating system core files (Linux, AIX and other supported platforms): crash signatures, failing modules, signal information and probable causes, without manual debugger work.
- Virtual machine and hypervisor data: resource starvation inside the guest, host contention, overcommitted CPU or memory, and storage latency.
- Application, web server, middleware and OS logs: other structured and semi-structured diagnostics, processed by the same reasoning engine.
Support matrix
| Area | Input | How to submit | What DrWatson looks for |
|---|---|---|---|
| Informix | Assertion Failure (AF) files | Single af.<hex> file, or inside an archive | Engine failures and internal errors (Health & Errors) |
| Informix | onstat -a output | onstat-a.out, onstat-a or onstat_a, or inside an archive | Performance, locks, memory, I/O |
| Informix | ifxcollect data (all collection profiles) | Inside an archive | System state, workload, capacity |
| Informix | ONCONFIG | Inside an archive, usually from ifxcollect | Configuration Audit |
| Informix | Query plans (sqexplain.out) | Single sqexplain.out file | Query performance |
| WebSphere | WebSphere logs | Inside an archive | Hangs, configuration issues |
| JVM | Thread dumps and javacores | Inside an archive | Hangs, deadlocks, thread contention |
| JVM | Heap dumps | Inside an archive | Memory leaks, excessive GC |
| Operating system | Core files | Inside an archive | Crash signatures, failing modules, signals |
| Operating system | OS logs | Inside an archive | System-level errors and events |
| Virtualisation | Guest VM logs and metrics | Inside an archive | CPU, memory and I/O starvation inside the VM |
| Virtualisation | Hypervisor host logs and performance data | Inside an archive | Host contention, overcommit, datastore and storage latency |
| Other | Application, web server and middleware logs | Inside an archive | Application and middleware errors |
Accepted archive formats: .tar, .tar.gz,
.tgz, .tar.bz2, .tbz2, .tar.zst,
.tzst and .zip.
What a DrWatson RCA report gives you
- Root-cause findings, prioritised by severity
- Recommendations ready to act on
- The evidence trail behind every conclusion
- Reports for different audiences: administrators, architects, management and executives
- Consistent diagnoses that keep expert knowledge in the organisation and shorten MTTR
Turn diagnostic evidence into answers
See what DrWatson finds in your systems. Arrange a technical demonstration, or start with a free analysis. Read Security & Privacy first to see how your data is handled.
See how DrWatson helps DBAs and architects.