kamel run simple.js --dev hangs after multiple kill/start

[aldettinger]

How to reproduce:
kamel run simple.js --dev
wait for the example to start successfully
CTRL+C

Repeat the 3 steps above 3 times (perhaps more), then it occurs that the example stay stuck with log:
Condition "Ready" is "True" for Integration simple
No integration pods looks to be running

[lburgazzoli]

can you check if there is any unsatisfied condition in the deployment ?

[aldettinger]

oc get deployment simple -o json output below status:

    "status": {
        "availableReplicas": 1,
        "conditions": [
            {
                "lastTransitionTime": "2020-06-10T09:18:39Z",
                "lastUpdateTime": "2020-06-10T09:18:39Z",
                "message": "Deployment has minimum availability.",
                "reason": "MinimumReplicasAvailable",
                "status": "True",
                "type": "Available"
            },
            {
                "lastTransitionTime": "2020-06-10T09:18:36Z",
                "lastUpdateTime": "2020-06-10T09:18:39Z",
                "message": "ReplicaSet \"simple-657f4555dd\" has successfully progressed.",
                "reason": "NewReplicaSetAvailable",
                "status": "True",
                "type": "Progressing"
            }
        ],
        "observedGeneration": 1,
        "readyReplicas": 1,
        "replicas": 1,
        "updatedReplicas": 1
    }

But, the status looks to be the same whether the example runs fine or hung up.
pod and integration statuses also remain equivalent.

[tadayosi]

I seem to reproduce the issue, too. I use Camel K 1.1.0-SNAPSHOT and minikube v1.11.0.

$ kamel run --dev examples/simple.js
...
Condition "DeploymentAvailable" is "True" for Integration simple: deployment name is simple
Condition "ServiceAvailable" is "False" for Integration simple: no http service required
Condition "ExposureAvailable" is "False" for Integration simple: no host or service defined
Condition "Ready" is "False" for Integration simple
Condition "Ready" is "True" for Integration simple

[tadayosi]

What hangs actually is the kamel run --dev session that's supposed to keep output logs. The deployment still runs fine as you can see running logs from another terminal with kamel log simple.

[tadayosi]

@aldettinger @lburgazzoli If no one is working on it yet, I can look into it.

[aldettinger]

Hey @tadayosi , sure you can 👍

[aldettinger]

@tadayosi it seems I can't reproduce with Camel-K 1.0.1, maybe linked to #1558

[tadayosi]

@aldettinger Yes, thanks to #1558 now cleanup of terminating pods seems to get faster. I think that's the reason why the issue no longer happens as frequently as before (less concurrent reads/writes of the pipe to stdout). I still have a feeling that the root cause is some deadlock at the shared pipe to stdout in the SelectorScraper. But for now the issue becomes negligible so let's close it.