mirror of
https://github.com/OPCFoundation/UA-.NETStandard.git
synced 2026-10-18 06:31:35 +08:00
romanett
d623007a3f
# Description Removes the Azure Pipelines CI and package-publishing definitions from `master`. GitHub Actions already owns all CI and package publishing for `master`, but pushes to `master` still queued the Azure definitions. They duplicated the Actions work and posted failing checks, for example [`Nuget Preview (Push Nuget Preview Pack Nugets Debug)`](https://github.com/OPCFoundation/UA-.NETStandard/runs/109065914524) on `19dc504c5`. ## Removed - `azure-pipelines.yml`, `azure-pipelines-preview.yml` - `.azurepipelines/` templates: `ci.yml`, `coverage.yml`, `preview.yml`, `sample-ha-longhaul.yml`, `sln.yml`, `test.yml`, `test-aot.yml` - Helpers only those templates used: `get-matrix.ps1`, `get-root.ps1`, `evaluate-test-results.ps1` - `TrxGateScriptTests`: it tested only `evaluate-test-results.ps1`. The Actions equivalent, `.github/scripts/get-test-verdict.ps1`, keeps its own `CiTestVerdictTests`. ## Kept **OneFuzz:** `.azurepipelines/onefuzz.yml` stays. Submitting to the internal OneFuzz service needs the Azure DevOps-only `onefuzz-task`, which has no GitHub Actions equivalent. The pipeline has no push, PR or schedule trigger, so it never runs on `master` pushes. The public drop build keeps running in the GitHub `fuzz-public-drop` job. **Scripts GitHub Actions still invokes:** these stay in `.azurepipelines/`, so every workflow path and `release.yml`'s candidate-bound checkout keep working. Only `generate-slnx.ps1` changes, and only its comment. | Script | Used by | | --- | --- | | `set-version.ps1` | buildandtest, nightly, nuget-publish, docker-image, stress / stability / long-haul, `run-dotnet-tests` action | | `generate-slnx.ps1`, `signlist{Debug,Release}.txt`, `validate-source-generator-packages.ps1` (+ `expected-packages.txt`) | nuget-publish | | `package-version-policy.ps1`, `validate-nuget-package-set.ps1`, `validate-preview-package-ordering.ps1`, `assert-published-packages-match.ps1` | nuget-publish, release | | `check-coverage.ps1` | buildandtest, nightly | | `validate-migration-plugin.ps1` | buildandtest | I didn't rename the directory on purpose. `release.yml` checks out the candidate commit and runs `./.azurepipelines/package-version-policy.ps1` from it, so moving the scripts would break promotion of any candidate built before the move. ## Docs and comments - `UA.slnx` drops the deleted files. - `codecov.yml`, `coverage-thresholds.json`, `targets.props`, the Redundancy sample test comments, the `.github` scripts and workflows, `CONTRIBUTING.md`, `docs/` and `fuzzing/` no longer describe Azure as running on `master` (`/azp run`, the `azure` Codecov flag, "the Azure gate", the Azure long-haul and AoT legs). - In the Developer Guide, the "Triggering a pipeline run" and "Migration status" sections are replaced by "Pull requests from outside contributors" and "Azure Pipelines on other branches". - The new section names `onefuzz.yml` as the one remaining Azure definition. - It keeps the warning that the `Release` ruleset's Azure required check must be lifted before cutting a new `release/2.<minor>` branch. - `master` is merged in, including the #4555 docs rewrite. Its new `docs/RedundancySampleTests.md` needed the Azure references removed again. `docs/README.md` already had none. ## Reviewer notes - **Azure DevOps side (admin action, not in this PR):** the definitions are configured in the Azure DevOps portal. With the YAML gone, `master` pushes give the CI and preview definitions nothing to run. Disabling definition 14's push trigger/schedule, definition 16's build-completion trigger, and the stale definition 13 stops any residual "file not found" runs. Leave the OneFuzz definition enabled. - `master378`, the 1.x `release/*` lines and the frozen `release/2.0.0` keep their own `azure-pipelines.yml`. This PR only touches `master`. ## Related Issues - Follow-up to #4513 (GitHub Actions for CI and nightly). ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [ ] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [ ] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [ ] I have addressed **all** PR feedback received. This PR removes a test and adds none. Locally, at `9b00da817`, the full `Opc.Ua.Tools.Tests` suite passed on net10.0: 884 passed, 1 skipped. The later commits (OneFuzz restore, merge with `master`) touch only YAML and docs. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
79 lines
3.9 KiB
Markdown
79 lines
3.9 KiB
Markdown
# Redundant Sample Integration Tests
|
|
|
|
The `Opc.Ua.Redundancy.Samples.Tests` project runs process-level integration
|
|
tests against the `RedundantServer`, `RedundantClient`, and `RedundantPubSub`
|
|
sample applications. The tests launch the applications as external processes
|
|
and assert on logged high-availability events such as `FAILOVER:`, `DATA LOSS:`,
|
|
`HA OK:`, and reconnect messages. They cover supported sample configurations,
|
|
catch regressions in end-to-end failover, and demonstrate when failover loses
|
|
data.
|
|
|
|
The test project includes **short-haul** tests for normal pull-request
|
|
validation and **long-haul** soak tests for extended runs on dedicated CI jobs.
|
|
|
|
## Contents
|
|
|
|
- [Short-haul tests](#short-haul-tests-pull-request-validation)
|
|
- [Long-haul tests](#long-haul-tests-manual--scheduled-soak)
|
|
- [Multi-replica leader-election failover](#multi-replica-leader-election-failover)
|
|
|
|
## Short-haul tests (pull-request validation)
|
|
|
|
Short-haul tests are deterministic and complete in seconds. The standard
|
|
[CI](../.github/workflows/buildandtest.yml) workflow discovers and runs them because the project follows the
|
|
`tests/Opc.Ua.*.Tests` naming convention. They use the NUnit category
|
|
`SampleHaShortHaul`:
|
|
|
|
* **PubSub demo, hot mode** — runs `RedundantPubSub --role demo --ha-mode hot` and asserts the promoted publisher continues the SequenceNumber
|
|
across failover with no reset (`SIMULATED: HA OK: sequence continued ...`).
|
|
* **PubSub demo, cold mode** — runs the same demo in cold mode and asserts the SequenceNumber reset (data loss) is made visible
|
|
(`SIMULATED: DATA LOSS: sequence reset ...`).
|
|
* **Server + client connectivity** — launches a `RedundantServer` and a `RedundantClient` together and asserts the client connects and begins
|
|
high-availability monitoring.
|
|
|
|
Run them locally from the repository root:
|
|
|
|
```powershell
|
|
dotnet test tests/Opc.Ua.Redundancy.Samples.Tests/Opc.Ua.Redundancy.Samples.Tests.csproj --filter "Category=SampleHaShortHaul"
|
|
```
|
|
|
|
## Long-haul tests (manual / scheduled soak)
|
|
|
|
Long-haul tests use `[Explicit]` and the NUnit category
|
|
`SampleHaLongHaul`, so they do not run on pull requests. They loop for a
|
|
configurable duration and check high-availability behavior across many
|
|
failovers:
|
|
|
|
* **PubSub failover soak** — repeatedly runs the PubSub demo in hot and cold modes for the whole duration, asserting the correct continuity
|
|
(hot) and data-loss (cold) narrative every iteration.
|
|
* **Client reconnect soak** — runs a `RedundantClient` against a
|
|
`RedundantServer` for the entire duration. The test repeatedly stops and
|
|
restarts the server, then checks that the client detects the outage,
|
|
reconnects when the server returns, and does not crash.
|
|
|
|
The `SAMPLE_HA_DURATION_MINUTES` environment variable controls the soak
|
|
duration (default: `60` minutes). Run a short local soak with:
|
|
|
|
```powershell
|
|
$env:SAMPLE_HA_DURATION_MINUTES = "5"
|
|
dotnet test tests/Opc.Ua.Redundancy.Samples.Tests/Opc.Ua.Redundancy.Samples.Tests.csproj --filter "Category=SampleHaLongHaul"
|
|
```
|
|
|
|
The long-haul tests run in CI through the [Sample HA Long-Haul Test](../.github/workflows/sample-ha-longhaul.yml)
|
|
GitHub Actions workflow (`workflow_dispatch` with a `duration` input, plus a weekly schedule).
|
|
|
|
## Multi-replica leader-election failover
|
|
|
|
Full multi-replica leader-election topologies include strong (Raft)
|
|
active/passive and eventual (CRDT gossip) active/active configurations. They
|
|
use a stable virtual endpoint with DNS-based re-resolution to a surviving
|
|
replica.
|
|
|
|
The Docker Compose setups under
|
|
[`samples/Redundancy/RedundantServer`](../samples/Redundancy/RedundantServer)
|
|
and [`samples/Redundancy/RedundantClient`](../samples/Redundancy/RedundantClient)
|
|
run several replicas on a container network. Stop the active replica to
|
|
observe cross-replica failover. The process-level long-haul soak above focuses
|
|
on deterministic failover detection, reconnect, and data-loss visibility
|
|
without container networking.
|