tft-request-watch 0.36.0

Wait on Testing Farm requests; restart on retriable infrastructure errors

tft-request-watch

Wraps Testing Farm CLI with unattended request supervision.

This package can be found

• on Copr as @cki/tft-request-watch
• on PyPI as tft-request-watch

The Copr packages are provided for all active Fedora releases and EPEL 10.

The code is tested on the supported Python versions 3.11, 3.12 and 3.13.

Benefits compared to running testing-farm watch directly:

• Keep polling through temporary API outages.
• Enforce a hard timeout with automatic cancellation.
• Automatically retry the whole request when known retriable infrastructure failures appear in result.summary.
• Restart only the affected plan(s) when a required test task is aborted due to an infrastructure issue.

Install

# from Copr
sudo dnf copr enable @cki/tft-request-watch
sudo dnf install tft-request-watch

# from PyPI
pip install tft-request-watch

# from source
pip install git+https://gitlab.com/redhat/centos-stream/tests/kernel/tft-request-watch

Usage

tft-request-watch provides two subcommands: watch and merge.

watch

Wait for a Testing Farm request to finish, restarting on retriable failures.

tft-request-watch watch --request-id YOUR-REQUEST-UUID

The default API base is https://api.dev.testing-farm.io/v0.1; use --api-url to point elsewhere.

Key options:

Option	Description
--request-id UUID	Request UUID to watch.
--dry-run	Skip write operations; log what would have been done.
--result-file PATH	Write per-run results JSON to PATH (needed for merge).
--required-task TASK	Restart if task is aborted; see below. Repeatable.
--max-restarts N	Maximum restart iterations (default: 3).
--hard-timeout-hours H	Cancel request after H hours (default: 24).

--required-task TASK restarts affected plans when a task result is error or pending in the XUnit output. Use the test path without the discover alias prefix (e.g. /distribution/install). Can be specified multiple times.

See tft-request-watch watch --help for all options including HTTP retry and timeout tuning.

Result file format

When --result-file is given, a JSON list is written with one entry per run (original + any restarts):

[
  {"id": "uuid-of-original-run", "xunit_url": "https://...", "...": "..."},
  {
    "id": "uuid-of-retry",
    "retry_of": "uuid-of-original-run",
    "restarted_plans": {"x86_64": ["/plans/tier1"]},
    "xunit_url": "https://...",
    "...": "..."
  }
]

restarted_plans identifies which (arch, plan) testsuites in the original XUnit were replaced by the retry, useful for tft-request-watch merge.

merge

Produce a single merged XUnit XML from a --result-file JSON, replacing testsuites that were restarted with the retry results.

tft-request-watch merge result.json merged.xml

Fetches the XUnit XML for every run recorded in result.json. Testsuites annotated via restarted_plans are taken from the retry; all other testsuites are kept from the original run.

tft-request-watch merge --help

Global options

Option	Description
--debug	Enable DEBUG logging on stderr.
--version	Print version and exit.

Development

Installing development dependencies:

pip install -e . --group dev

Running linting/tests:

tox

Versioning

This project uses semver. The major version is maintained in tft_request_watch/__init__.py. Minor versions are automatically generated on each push to main using the GitLab pipeline ID, producing versions like 0.42.0, 0.43.0, etc.

Bump the major version for breaking changes.

License

GPLv2+ — see LICENSE.