Failing test(s): TestAccDataprocCluster_withNodeGroupAffinity

[SarahFrench]

• Failure rate: 38% since 2023-05-03

Impacted tests:

• TestAccDataprocCluster_withNodeGroupAffinity

Affected Resource(s)

• google_dataproc_cluster

Nightly builds:

• Test History: TestAccDataprocCluster_withNodeGroupAffinity

Message:

provider_test.go:307: Step 1/1 error: Error running apply: exit status 1
Error: Error waiting for creating Dataproc cluster: Error code 9, message: Instance could not be scheduled due to no matching node with property compatibility.
Explanation:
The matching node(s) <test-nodegroup-wk1cg8gvhx-252j, test-nodegroup-wk1cg8gvhx-t8g5, test-nodegroup-wk1cg8gvhx-z6bc> are not in READY status.
Potential fix:
- Please consider waiting for the matching node(s) to become READY.
- Please consider deleting and recreating the node or contacting Google Cloud Support, in case it is taking several minutes for the node to become READY.
with google_dataproc_cluster.basic,
on terraform_plugin_test.tf line 25, in resource "google_dataproc_cluster" "basic":

b/299683841

[SarahFrench]

This test is also affected by quota limits, possible exacerbated by increasing the # of parallel tests

[melinath]

This seems to be flipping between quota issues and the other issue. I'm going to keep this as a service/dataproc issue; if it turns out we need to increase quota we can do that later.

[melinath]

Possibly the node groups here should also use a tf-test prefix.

[melinath]

Currently 50% failure - 62 / 124.

[roaks3]

@melinath I ran the labeler today and the service/dataproc label was re-added because of the Affected Resource(s) block. It looks like the service/compute-sole-tenancy label was added at one point because the formatting was off such that the rest of the description was included as Affected Resource(s), and google_compute_node_group was found.

I think we want to keep service/dataproc and remove service/compute-sole-tenancy, but wanted to double check if there was a reason for removing service/dataproc previously.

[melinath]

I must have clicked on the wrong label? +1 that this is clearly dataproc, not compute-sole-tenancy

[NickElliot]

to amend this ticket, the most common test error message is the following, the quota issue being relatively rare:

        Error: Error waiting for creating Dataproc cluster: Error code 9, message: Instance could not be scheduled due to no matching node with property compatibility.

        Explanation:
        The matching node group(s) <test-nodegroup-randomsuffix> do not match the intance's machine family type.