Yarik

Data race at interactive feature tests

1

Noticed in test-multiuser-ubuntu test: https://community-tc.services.mozilla.com/tasks/RslJcW8fTLykFJct853Oqw/runs/0/logs/public/logs/live.log ``` ================== WARNING: DATA RACE Read at 0x00c00007e4c8 by goroutine 15: github.com/taskcluster/taskcluster/v56/workers/generic-worker/interactive.(*InteractiveJob).Terminate() /task_169831511858264/taskcluster/workers/generic-worker/interactive/interactivejob.go:80 +0x5b github.com/taskcluster/taskcluster/v56/workers/generic-worker/interactive.(*InteractiveJob).handleWebsocketMessages() /task_169831511858264/taskcluster/workers/generic-worker/interactive/interactivejob.go:127 +0x5c8 github.com/taskcluster/taskcluster/v56/workers/generic-worker/interactive.CreateInteractiveJob.func3() /task_169831511858264/taskcluster/workers/generic-worker/interactive/interactivejob.go:74 +0x33 Previous write at 0x00c00007e4c8 by...

UI: Full task definition page doesn't have margins

8

For example: https://dev.alpha.taskcluster-dev.net/tasks/OsfAluRnTca6e5LEZv_mew/definition There is no gap on the left making it harder to read

Kubernetes logs show high number of "readiness probe failed"

1

[gcp error](https://console.cloud.google.com/errors/detail/CJnOtciylYOLeQ;filter=%5B%5D;time=P1D?project=moz-fx-taskcluster-prod-4b87) Some of the errors are seen on pod creation, which means that `initialDelaySeconds: 3` might be too small, as it takes longer to start (~10s). Might be also...

While debugging connections in profiler I noticed that node spends a huge amount of time loading `googleapis` package. That package (as of 73.0.0) contains ~218 service that take about 84Mb....

at the moment there are no `kubectl --dry-run=client` in our testing pipelines, so wrong configration can pass tests. We should add dry run to avoid this.

Large amount of "duplicate key value violates unique constraint" in db logs

5

While investigating why queue artifacts might not be expiring, I noticed that there are many similar db errors in logs: ``` 2023-04-06 08:53:49.889 UTC [1517417]: [13-1] db=taskcluster,user=taskcluster_queue ERROR: duplicate key...

I was going to add test for the async client, but noticed that there are no test fixtures set up there to make it easy, so decided to skip it....

Update integration guidelines documentation

2

This page https://docs.taskcluster.net/docs/manual/using/integration/frontend includes an example code that no longer works. It references old versions of `hawk` and `taskcluster-client-web` that no longer work or exist. Todo: provide up-to date documentation...

We see in logs following message, that was observed since 55.1.0 (according to GCP) ``` superagent: double callback bug ``` ``` { code: "ETOOLARGE" message: "Maximum response size reached" name:...

Tasks with dependencies might not be picked up by workers before deadline

1

Happens on Community-TC Tasks from and often gets resolved with exception `deadline-exceeded` Some examples: [ECeRNdylQcGeslW5DZUvJw](https://community-tc.services.mozilla.com/tasks/groups/ECeRNdylQcGeslW5DZUvJw), [NCTxhSL_RDSy_FtYNSPBwA](https://community-tc.services.mozilla.com/tasks/groups/NCTxhSL_RDSy_FtYNSPBwA) Observations: * Those task that fail have dependencies. * All running in `proj-relman/generic-worker-ubuntu-22-04` workers...