Update Spin and Go versions for actions

[kate-goldenring]

We are seeing some CI errors seemingly due to too old of Go and TinyGo. This also bumps Spin version in CI.

[kate-goldenring]

@Mossaka I am a little lost at why the tests are failing. I can build the images locally. The cluster logs show that the spin app is having a hard time connecting with the MQTT broker (but there are failures in HTTP trigger apps too):

time="2024-10-29T04:01:23.914095692Z" level=info msg="StartContainer for \"83894a26ccd9b450c7ba18df23b946c5d7214d797c1e85ae570ec7cf69f5a2f5\" returns successfully"
time="2024-10-29T04:01:24.820250695Z" level=error msg="run_wasi ERROR >>>  failed: failed to connect to 'mqtt://10.43.174.211:1883'

Caused by:
    [-1] TCP/TLS connect failure"

I wonder if there is an issue with the k3d cluster certs

[kate-goldenring]

new hypothesis: the tests are failing because the runner is running out of space to execute the containers. Testing by no longer using an in cluster EMQX broker (250+ MB) and instead using test.mosquitto.org with a very unique topic

Message:          The node was low on resource: ephemeral-storage. Threshold quantity: 3892562797, available: 3696304Ki. Container redis was using 36Ki, request is 0, has larger consumption of ephemeral-storage. 

[kate-goldenring]

MQTT test without local broker is passing but we are still failing to get all apps running:

NAME                                        READY   STATUS    RESTARTS   AGE    IP          NODE                        NOMINATED NODE   READINESS GATES
spin-outbound-redis-584db675d7-ljlxt        1/1     Running   0          2m     10.42.2.6   k3d-test-cluster-server-0   <none>           <none>
spin-static-assets-67b7b5878b-7ztfx         1/1     Running   0          2m     10.42.0.7   k3d-test-cluster-agent-0    <none>           <none>
wasm-spin-5f4cfb8886-5q4kf                  1/1     Running   0          2m     10.42.0.5   k3d-test-cluster-agent-0    <none>           <none>
spin-keyvalue-7995f66b8d-mp9d5              1/1     Running   0          2m     10.42.0.6   k3d-test-cluster-agent-0    <none>           <none>
spin-mqtt-message-logger-6b57c956c5-5546v   1/1     Running   0          2m     10.42.1.6   k3d-test-cluster-agent-1    <none>           <none>
spin-multi-trigger-app-7f7586b9b-5dzmb      0/1     Error     0          2m     10.42.1.5   k3d-test-cluster-agent-1    <none>           <none>
redis                                       1/1     Running   0          2m1s   10.42.2.5   k3d-test-cluster-server-0   <none>           <none>

This time it is the redis trigger component (which is a part of spin-multi-trigger-app-7f7586b9b-5dzmb). There is likely a race case where the spin app is starting before the redis one based on age ^^

[Mossaka]

here is likely a race case where the spin app is starting before the redis one based on age ^^

Interesting, maybe we should wait longer to allow k8s to restart failed containers?

[Mossaka]

I'd prefer we keep IDE specific settings to local computer, not upstream.

[kate-goldenring]

Go evicted pods again do to resource constraints. Trying again with only one agent in k3d cluster. If this doesn't resolve it, i think we should look into if GH runners reduced their default size. Im confused by the suddenness of this. I wont be able to get back to this for a couple days so fingers crossed this does the trick

[kate-goldenring]

We are still getting disk pressure:

Reason:              Evicted
Message:             Pod was rejected: The node had condition: [DiskPressure]. 

[kate-goldenring]

Next test is modifying the kubelet eviction threshold https://k3d.io/v5.4.2/faq/faq/#pods-evicted-due-to-lack-of-disk-space

[kate-goldenring]

@Mossaka @devigned @radu-matei @jsturtevant can I have one more look at this? Finally got CI passing by doing to rust and docker cleanup before we run the tests