Note: tests pass with the following patch, which I'm not sure I should apply?

diff --git a/nixos/tests/cockroachdb.nix b/nixos/tests/cockroachdb.nix
index d0cc5e19837..9ea271fb49e 100644
--- a/nixos/tests/cockroachdb.nix
+++ b/nixos/tests/cockroachdb.nix
@@ -92,9 +92,6 @@ let
 
       # Hold startup until Chrony has performed its first measurement (which
       # will probably result in a full timeskip, thanks to makestep)
-      systemd.services.cockroachdb.preStart = ''
-        ${pkgs.chrony}/bin/chronyc waitsync
-      '';
     };
 
 in import ./make-test-python.nix ({ pkgs, ...} : {

@anmonteiro it seems this was once necessary, as cockroachdb was very sensitive to slightly differing clocks. Did you read the long comment section in nixos/tests/cockroachdb.nix?

On master, I tried removing all the chrony parts and immediately was able to get nodes kicked out of the cluster:

clock synchronization error: this node is more than 500ms away from at least half of the known nodes (0 of 1 are within the offset)

I didn't try on your PR. Are you aware of any changes done in cockroachdb 20.x regarding that?

@flokli thanks for the suggestion. I just tried removing all the chrony parts and I do get the same behavior that you describe, suggesting nothing changed in CockroachDB v20.1.

Let me restate my comment in the original post that opened the PR, as it could probably be clearer: with the current test definitions, I'm suspicious that chronyc waitsync never finishes, which leads to a timeout starting up any CockroachDB node and therefore running the tests. Removing that preStart command made it work, and tests passed in that configuration.

I'm trying other alternative commands to make tests pass. In the meantime, I'm open to other suggestions.

Ok, so some more investigation revealed that the chronyd service in the VM just doesn't start because it's missing /dev/ptp0 and there's a ConditionPathExists for /dev/ptp0.

It sounds like I just can't run the tests locally, so I'd appreciate someone trying it out. This setup is working for me locally, while I'm developing against CockroachDB, but I have admittedly not tried the multi-node setup.

@thoughtpolice can you give this a look, especially w.r.t. #51338?

This has been mostly superseded by a bump to 20.1.4. This now only adds the license change, which we're still missing.

I just rebased this on master, and now the diff only contains the license change.

This does not seem to be rebased, but merged - now containing 3 commits. Can you rebase this, so it's a single commit changing the license?

@flokli I just squashed the commits as you suggested.