Tw continuous ready match

[scott-cotton]

No description provided.

[gitar-bot]

Gitar automatically fixes CI failures and addresses comments starting with Gitar.

⚙️ Options:

• Fix CI failures*

🔄 To revert changes, post a comment:

Gitar revert to commit <desired commit SHA>

---

📚 Docs

* Gitar never force pushes to your branch

[daniel-de-vera]

@scott-cotton, here is another issue:

$ signadot sandbox apply -f - <<EOF
name: test1
spec:
  cluster: test
  virtual:
  - name: location
    workload:
      kind: Deployment
      namespace: hotrod-devmesh
      name: location
  routing:
    forwards:
      - name: test
        port: 8081
        toLocal: "localhost:18081"
EOF

$ signadot traffic watch --sandbox test1 --dir /tmp/tw1
Error: 400 Bad Request: sandboxes with spec.local or spec.routing.forwards must be created via the Signadot CLI

[scott-cotton]

@daniel-de-vera I merged the addition of doneAt here for workflow reasons.

It contains fixes for your comments above, in addition to @foxish 's suggestion to overwrite on first apply.

It does not yet contain adding local machine id when needed.

[daniel-de-vera]

$ signadot traffic watch --sandbox test1 --short
Waiting (up to --wait-timeout=30s) for sandbox to be ready...
✓ Sandbox status: Ready: All desired workloads are available.
msg="watching sandbox request activity" sandbox=test1 watch-options=-trunc(0)/-trunc(0) output=<none>


msg=incoming-request sandbox=test1 request.id=d4404429 request.normHost=location.hotrod-devmesh.svc.cluster.local:8081 request.dest=Deployment/hotrod-devmesh/location request.uri=http://location.hotrod-devmesh.svc:8081/locations request.method=GET request.userAgent=curl/8.5.0
msg=request-done sandbox=test1 request.id=d4404429 request.doneAt=2025-10-14T14:08:41.42059431-03:00


msg=request-done sandbox=test1 request.id=2dcf3f81 request.doneAt=2025-10-14T14:08:46.02429807-03:00
msg=incoming-request sandbox=test1 request.id=2dcf3f81 request.normHost=location.hotrod-devmesh.svc.cluster.local:8081 request.dest=Deployment/hotrod-devmesh/location request.uri=http://location.hotrod-devmesh.svc:8081/locations request.method=GET request.userAgent=curl/8.5.0


msg=request-done sandbox=test1 request.id=909417b3 request.doneAt=2025-10-14T14:08:51.290361624-03:00
msg=incoming-request sandbox=test1 request.id=909417b3 request.normHost=location.hotrod-devmesh.svc.cluster.local:8081 request.dest=Deployment/hotrod-devmesh/location request.uri=http://location.hotrod-devmesh.svc:8081/locations request.method=GET request.userAgent=curl/8.5.0

It's odd to find the request-done event in stdout before the incoming-request for the same request.id.
Do you think there could be a simple solution to this?

[scott-cotton]

as mentioned previously, no.

[scott-cotton]

I have however stacked #223 to address this.

I think it technically removes the limited memory guarantee, unfortunately, but in practice it would be near impossible for that to become an issue.

I also think it solves a race that would be hard to witness, being quite unlikely but technically possible, where the updated meta.json file could have a doneAt addition attempted before the initial write. I don't think that's practically likely b/c it waits for the request/response payload files to be written. But, technically the race I believe is there without the proposed change.

[daniel-de-vera]

LGTM