Enable zebrad sync tests by default

[teor2345]

Motivation

In the past, we have accidentally merged PRs that break zebrad's sync functionality. So we want to add a basic genesis sync test to the zebrad acceptance tests.

Solution

• enable the sync tests by default
• set the ZEBRA_SKIP_NETWORK_TESTS env var on any CI jobs that fail
  • Disable CI / Test on ubuntu-latest, because it fails
  • Check that other CI runners succeed
  • Check if Google Cloud Build fails
    • needs branch in ZcashFoundation/zebra? - see Test PR: Enable sync tests by default #1180
  • Check if tests for cd.yml fail
    • needs merge to main?

Related Issues

Closes #1141.

Follow-up work

CI Reliability

Make sure CI is reliable, even if the seeder sends us slow peers

• Tune the sync test timeouts, if needed
• Add sync retries, if needed

This work doesn't need a ticket - we only need to do it if we see CI failures.

Extended Sync Tests

Once we are sure that the genesis sync is reliable, we can add longer sync tests.
See #1004 (RFC) and #745 (tracking issue) for details.

[teor2345]

@dconnolly how can I test if this PR makes Google Cloud CD tests fail?
(The tests in this PR depend on the network configuration on the CI runner.)

This PR is from my GitHub repository, and #1180 is from a branch in the ZcashFoundation repository.

[teor2345]

The existing sync, state, and network code is unreliable, see #1181, #1182, and #1183.

So we might see some errors downloading genesis, but it's just one block, so the errors should not be too frequent.

[teor2345]

The Windows sync test fails intermittently, so I'm going to disable it.
Either Windows intermittently has a slow network, or some CI runners don't have network at all, or we're seeing the sync hang in #1181.

[dconnolly]

@dconnolly how can I test if this PR makes Google Cloud CD tests fail?
(The tests in this PR depend on the network configuration on the CI runner.)

This PR is from my GitHub repository, and #1180 is from a branch in the ZcashFoundation repository.

All tests in the Google Cloud Build are passing:

Can you load this url? This is the raw console out .

[teor2345]

The Windows sync test fails intermittently, so I'm going to disable it.
Either Windows intermittently has a slow network, or some CI runners don't have network at all, or we're seeing the sync hang in #1181.

I added #1189 to follow up, because we really should be testing sync on every platform. (Particularly Linux, if we want it to be our distribution platform.)

[teor2345]

Can you load this url? This is the raw console out .

Yeah, that works for me.

I guess I don't know exactly how Zebra CI works:

• is the some CI/CD that only runs on branches in ZcashFoundation/zebra?
• is there some CI/CD that only runs on merges to main?

[dconnolly]

Can you load this url? This is the raw console out .

Yeah, that works for me.

I guess I don't know exactly how Zebra CI works:

• is the some CI/CD that only runs on branches in ZcashFoundation/zebra?
• is there some CI/CD that only runs on merges to main?

Oh, yes:

• ci.yml is on PRs
• cd.yml is on main
• coverage.yml is both

[dconnolly]

Can you load this url? This is the raw console out .

Yeah, that works for me.
I guess I don't know exactly how Zebra CI works:

• is the some CI/CD that only runs on branches in ZcashFoundation/zebra?
• is there some CI/CD that only runs on merges to main?

Oh, yes:

• ci.yml is on PRs
• cd.yml is on main
• coverage.yml is both

But cd and ci both build the same Dockerfile in gcloud, which has the same commands, just one is via a gcloud action and one is via the Google Cloud Build App integration to allow it to work on PRs from external forks.

[teor2345]

* ci.yml is on PRs
* cd.yml is on main
* coverage.yml is both

Ok, so here's our current test matrix for the sync tests:

• macOS CI on pull requests
• Debian Linux CD on main branch merges

(The other CI runners have unreliable networks.)

I think that's enough for now.