roachprod: classify dns errors

[herkolategan]

Previously DNS errors were not marked making it hard to classify these errors
later on. This change marks all DNS related errors in order to better report on
the errors.

roachtest then classifies these errors as infra flake to help with triaging.

Epic: None
Release Note: None

[cockroach-teamcity]

This change is 

[smg260]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @DarrylWong, @herkolategan, @rachitgsrivastava, and @renatolabs)

---

pkg/cmd/roachtest/github.go line 155 at r2 (raw file):

		messagePrefix = fmt.Sprintf("test %s was skipped due to ", testName)
		infraFlake = true
	case failureContainsError(firstFailure, rperrors.ErrSSH255):

These 2 cases can probably be collapsed into one. Might be good to add a test case for this. (There are already a bunch).

Also this will result in a brand new umbrella github issue for just dns_problem (if that's the way we want to go)

---

pkg/roachprod/vm/gce/dns.go line 271 at r1 (raw file):

// markDNSOperationError should be used to mark any external DNS API or Google
// Cloud DNS CLI errors as DNS operation errors.
func markDNSOperationError(err error) error {

nit: not sure this needs to be a separate function

[herkolategan]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @DarrylWong, @rachitgsrivastava, @renatolabs, and @smg260)

---

pkg/cmd/roachtest/github.go line 155 at r2 (raw file):

Previously, smg260 (Miral Gadani) wrote…

These 2 cases can probably be collapsed into one. Might be good to add a test case for this. (There are already a bunch).

Also this will result in a brand new umbrella github issue for just dns_problem (if that's the way we want to go)

Yes, we want a new umbrella issue, as intermittent network issues seem to be causing some failures and we want to track it. I like to have the cases explicit as it's easier to reference.
I'll add a test.

---

pkg/roachprod/vm/gce/dns.go line 271 at r1 (raw file):

Previously, smg260 (Miral Gadani) wrote…

nit: not sure this needs to be a separate function

Probably not, but I felt like it made wrapping the errors a bit cleaner and easier to see what's going on.

[smg260]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @DarrylWong, @rachitgsrivastava, and @renatolabs)

[herkolategan]

TFTR!
bors r=smg260

[herkolategan]

bors r-

[craig]

Canceled.

[herkolategan]

bors r=smg260

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)