You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 25997cb
Browse filesBrowse the repository at this point in the historyBrowse files
|`mise run e2e:odh:python`| Supported upstream Python gateway E2E in the shared OpenShift mode |
208
209
|`cargo nextest run --config-file "$PWD/.config/nextest.toml" --tool-config-file "odh:$PWD/e2e/rust/tests/odh/nextest.toml" --profile e2e-odh --manifest-path e2e/rust/Cargo.toml --features e2e-odh --test odh -E 'test(=module::test_name)'`| A single ODH test function |
209
210
211
+
### Python gateway E2E
212
+
213
+
`e2e:odh:python` reuses the public ODH entrypoint: it deploys the shared
214
+
OpenShift gateway, imports the example provider profiles, runs upstream
215
+
`e2e/python`, and tears the deployment down. It writes the phase-scoped report
216
+
`e2e-odh-python-shared-python.xml` and merged `e2e-odh-python-xunit.xml` plus HTML
217
+
companion in `results/`. `odh/python-gates.toml` contains the exact tests that cannot run
218
+
on the current OpenShift lane (host-only fixtures, unrestricted egress, the
219
+
missing Python workload image, and the pending upstream deletion-timeout
220
+
decision). It does not run the Rust image-provenance check. Gate entries are
221
+
collection-validated so an upstream rename cannot
222
+
silently leave a stale exclusion. The phase also fails when the JUnit report
223
+
shows no executed test case, so a suite that skips every collected case cannot
224
+
report success. The Python tests also run as part of the tier2 shared phase,
225
+
riding the same gateway deployment.
226
+
227
+
The public `mise` tasks that include the Python runner (`e2e:odh:python` and
228
+
`e2e:odh:tier2`) provision `python:dev` first. If invoking
229
+
`odh/scripts/e2e-odh-entrypoint.sh` directly for one of those plan-backed
230
+
tiers, prepare that environment yourself with `mise run python:dev`; this is
231
+
intentional so the entrypoint fails before deploying when the test interpreter
232
+
is unavailable.
233
+
210
234
Example, running the Smoke tier against a real cluster:
211
235
212
236
```bash
@@ -306,10 +330,10 @@ mode. Phase JUnit files are named `e2e-odh-<tier>-<phase>.xml`; the entrypoint
306
330
merges them into `e2e-odh-<tier>-xunit.xml` for Jenkins discovery and
307
331
`e2e-odh-<tier>.html` in `results/`.
308
332
309
-
### Every `e2e:odh*` task runs the image provenance check
333
+
### Image provenance runs in every nextest-backed tier
310
334
311
335
`smoke::image_provenance::test_sandbox_gateway_supervisor_images` runs in the
312
-
shared phase of every logical tier. An empty phase fails instead of passing
336
+
shared phase of every nextest-backed tier. An empty phase fails instead of passing
313
337
with an empty report.
314
338
315
339
## Image provenance verification
@@ -413,7 +437,7 @@ locally with:
413
437
414
438
The local build tags the image as `odh-openshell-e2e`.
415
439
416
-
The image accepts `smoke`, `tier1`, `tier2`, `tier3`, `odh`, or `full`
440
+
The image accepts `smoke`, `tier1`, `tier2`, `tier3`, `odh`, `full`, or `python`
417
441
as its first argument; `smoke` is the default. The Shift-Left job mounts
418
442
cluster credentials and a writable report directory, then passes environment
reason = "requires Python or /sandbox/.venv in the sandbox; the current OpenShift workload is Ubuntu, while upstream builds e2e/python/Dockerfile.workload locally"
reason = "upstream Python E2E builds and selects e2e/python/Dockerfile.workload; this OpenShift lane uses the product-default Ubuntu workload, so these no-image sandbox tests never reach Ready"
reason = "the shared Python fixture's 30-second delete RPC deadline is shorter than the Kubernetes driver's bounded Pod-termination wait; an upstream fixture-timeout correction is being validated before these tests are re-enabled"
0 commit comments