Files
f6b27025a6 Classify sensitive log parameters with Microsoft.Extensions.Compliance data classification (#4864)
## Proposed changes

Fixes #4857. The stack's log messages wrote network addresses (endpoint,
discovery and reverse connect URLs, host names, IP addresses, ports),
application and user identities, file paths and exception texts as plain
parameters. Applications had no reliable way to keep them out of
general-purpose logs.

This PR replaces the stack's own redaction mechanism with the standard
.NET [data classification and
redaction](https://learn.microsoft.com/dotnet/core/extensions/data-redaction)
packages. Each sensitive `[LoggerMessage]` parameter is classified. The
application decides per classification what reaches its logs: erase it,
replace it with an HMAC token so entries stay correlatable, or pass it
through.

### Taxonomy
`Opc.Ua.Redaction.OpcUaDataClassifications` defines the `OpcUa` taxonomy
and one attribute per classification:

| Classification | Used for | Parameters |
|---|---|---|
| `[NetworkAddress]` | URLs, host and domain names, IP/MAC addresses,
end points, ports | 120 |
| `[ApplicationIdentity]` | certificates, subject, issuer, thumbprint,
ApplicationUri | 103 |
| `[FilePath]` | file, directory and certificate store paths | 67 |
| `[SensitiveText]` | exception or remote-supplied text passed as a
parameter | 55 |
| `[UserIdentity]` | user names, token subjects | 7 |

Correlation ids (`ChannelId`, `SessionId`, ...), status codes and
NodeIds stay unclassified.

### Pipeline and defaults
- Opc.Ua.Core now references
`Microsoft.Extensions.Telemetry(.Abstractions)` 10.11.0, so every
project above Core uses the `Microsoft.Gen.Logging` generator.
Opc.Ua.Client already did, through Http.Resilience. The packages ship
net462 and netstandard2.0 builds, so net48 is supported.
- **A plain `ILoggerFactory` logs `(null)` in place of classified
values.** Classified values are only written by the redaction-aware
logger.
- The new `ILoggingBuilder.EnableOpcUaRedaction()` turns that logger on.
It is `EnableRedaction()` with the discriminator off; otherwise every
unredacted value gets `:ParamName` appended.
`AddOpcUa().AddLogging(...)` and `DefaultTelemetry.Create(...)` call it,
as do the samples, so they log the same information as before until
redactors are registered.
- `services.AddOpcUa()` alone does not change the host's logging.
Generic-host apps call `builder.Logging.EnableOpcUaRedaction()`. This is
documented as a behavior-breaking change in
`docs/migrate/2.0.x/telemetry.md`.

### Obsolete API
`Redact`, `RedactionWrapper<T>`, `IRedactionStrategy`,
`RedactionStrategies` and `SimpleRedactionStrategy` are `[Obsolete]`.
They still work, but the stack no longer uses them.

### Not classified
- The `Exception` passed to a log call. Applications can drop exception
messages with `LoggerEnrichmentOptions.IncludeExceptionMessage`.
- Exception and `ServiceResult` message texts.
- `ServiceResult`/`StatusCode` parameters and logged data values.
- The few log calls whose level is decided at run time.
- `Opc.Ua.Vision.OpenUsd`, which references only Opc.Ua.Types and cannot
see the attributes.

### Generator behavior differences
These are visible to log consumers and to tests:
- `IFormattable` values such as `NodeId` are stored in the structured
state as invariant strings, not as objects.
- The generated log state is a thread-local object that is cleared after
the call.
- Classified values are pooled redactor objects, valid only during the
`Log` call.

Tests that inspected log state after the fact now capture it during the
call. The shared `RecordingLoggerProvider` snapshots classified values.

## Related Issues
- Fixes #4857

## Types of changes
- [ ] Bugfix (non-breaking change which fixes an issue)
- [x] Enhancement (non-breaking change which adds functionality)
- [ ] Test enhancement (non-breaking change to increase test coverage)
- [x] Breaking change (fix or feature that would cause existing
functionality to not work as expected; plain logger factories drop
classified values)
- [x] Documentation Update (if none of the other choices apply)

## Checklist
- [x] I have read the CONTRIBUTING doc.
- [x] I have signed the CLA.
- [x] I ran tests locally with my changes, all passed.
- [x] I fixed all failing tests in the CI pipelines.
- [x] I fixed all introduced issues with CodeQL and LGTM.
- [x] I have added tests that prove my fix is effective or that my
feature works and increased code coverage.
- [x] I have added necessary documentation (if appropriate).
- [ ] Any dependent changes have been merged and published in downstream
modules.

## Further comments
New tests:
- `LogClassificationCoverageTests` (Tools.Tests) scans all 104 stack
assemblies and fails if an address, identity or path parameter of a
`[LoggerMessage]` method is unclassified.
- `DataClassificationLoggingTests` (Core.Tests) pins erasing and HMAC
redaction, pass-through, the plain-factory drop, and both stack
defaults.

Local results on net10.0:

| Suite | Passed | Failed |
|---|---|---|
| Core.Tests | 6,489 | 0 |
| Client.Tests | 3,231 | 0 |
| Tools.Tests | 7,210 | 0 |
| PubSub.Tests | 1,536 | 0 |
| Core.Diagnostics.Tests | 571 | 0 |
| Configuration.Tests | 241 | 0 |

- **Server.Tests:** the full run predates the final test fixes and had
12 failures. All 12 were in the fixtures fixed afterwards, which now
pass: RoleStateBinding (45/45) and the FileDirectory and
compatibility-logging fixtures (74/74 in total). The full Core.Tests and
Server.Tests runs were not repeated after those fixes.
- **Builds:** `UA.slnx` builds with 0 errors for net10.0 and net48, and
the whitespace check passes.
- **NativeAOT:** not confirmed locally, because the machine has no C++
linker. A trimmed publish of `Opc.Ua.Aot.Tests` reported no trim or AOT
warnings from the stack, the new packages or the generated logging code;
its only errors come from the TUnit test framework (`FSharp.Core` not
found). The `aot-test` CI job has to confirm the native build.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Marc Schier <marcschier@users.noreply.github.com>
2026-10-21 09:47:07 +00:00
..

OPC 10000-110 Asset Management Basics samples

Two console applications that show the Opc.Ua.AMB* libraries end to end: a server that makes Device Integration devices and a Machinery machine manageable assets, and a client that finds and reads them with AmbClient.

Sample What it does
AmbServer Serves a cooling pump, the flow sensor it relies on and a hydraulic press as assets, and simulates their health alarms and maintenance
AmbClient Discovers every asset, prints a snapshot of each, the location hierarchies, adds and removes a documentation link, and streams health alarms and maintenance activities

The Asset Management Basics developer guide explains every API used here.

Run

In two terminals:

dotnet run --project samples/AMB/AmbServer -- --port 62553
dotnet run --project samples/AMB/AmbClient -- --insecure --seconds 60

--insecure makes the client accept the server's self-signed certificate; never use it in production. The client also takes --url <endpoint> (default opc.tcp://localhost:62553/AmbServer), --seconds <n> (how long to stream, 0 to skip) and --asset-id <id> (writes a new AssetId to the first asset that has one and finds the asset by it).

The server keeps what clients write - AssetIds, edited and added documentation links, the writable location and local time - below the local application data folder (OPC Foundation/AmbServer/state), or in the folder given with --state <directory>. Stop the server, start it again, and the client reads the values it wrote before.

What the server shows

File Shows
Program.cs AddMachinery() as the owner of the Device Integration address space, AddAssetManagement(...) with UseFileSystemStores, ConfigureAssetManagement(...) defining the location hierarchies and the ECLASS dictionary entry, ConfigureMachinery<T>() and ConfigureDevicesFor<MachineryNodeManager>(...)
Plant.cs Two DI devices registered with RegisterAsAssetAsync: every building block - configurable AssetId, DeviceHealth derived from the alarms, health alarms, maintenance activities with details, documentation links (fixed, editable, user links), version information, locations and local time, classification, requirements and capabilities, and a 0:Utilizes relation between the assets
PressConfigurator.cs A Machinery machine registered through IMachineHandle<T>.AsNode(); its servicing switches the machine into the Maintenance operation mode while it executes
PlantSimulation.cs A BackgroundService that raises and clears a bearing temperature alarm with the sensor as root cause once a minute, fails the self test of the sensor every five minutes (AssetRootCauses.Self), and moves a maintenance activity of the pump and of the press on every two minutes

Health alarms and maintenance activities are created while the owning node manager builds its address space, which is why Plant and PressConfigurator register the assets from the DI and Machinery setup.

What the client shows

Code in Program.cs Shows
connection.Session.AssetManagement(telemetry) Creating the client; registers the AMB structured types
amb.IsSupported Detecting whether the server publishes the AMB namespace
DiscoverAssetsAsync All assets from the AMB alias categories
ReadAssetAsync → AssetSnapshot Identification, DeviceHealth, alarms, maintenance, links, locations and classification in one pass
BrowseLocationsAsync The hierarchical and operational location trees and the assets in them
AddDocumentationLinkAsync, ReadDocumentationLinksAsync, RemoveDocumentationLinkAsync The DocumentationLinks AddIn
WriteAssetIdAsync, FindAssetsAsync Configuring an AssetId and finding the asset by it
ObserveHealthAlarmsAsync, AcknowledgeAsync Health alarms with severity, fault category and root causes; a cleared alarm is acknowledged
ObserveMaintenanceAsync Maintenance activities moving through Planned, Executing and Finished

SampleSession.cs is the session setup. It keeps the ApplicationInstance alive until the session closes, because the instance owns the CertificateManager the session validates with, and it creates a ManagedSession, whose default streaming subscription the observe calls use.

Troubleshooting

Symptom Fix
Client reports BadCertificateUntrusted Pass --insecure for the demo, or trust the server certificate in the client's PKI (under the local application data folder, OPC Foundation/AmbClient/pki)
AddLink / RemoveLink: refused: BadUserAccessDenied The server refuses anonymous link editing unless AuthorizeLinkEdit allows it; the sample server allows it
The stream shows nothing for a while The alarm becomes active 20 s into every minute and maintenance moves on every two minutes; stream longer with --seconds
Port already in use Start the server with --port <n> and the client with --url opc.tcp://localhost:<n>/AmbServer