Individually run tests succeed while running whole class fails on the second

[callaertanthony]

Describe the bug
I discovered when I run whole class test I have a second independent test failing when I need to create topics.

To Reproduce
This gist : https://gist.github.com/callaertanthony/571eeeca9b406b2cf0770df6394948bc

Expected behavior
The two tests must success as they run exactly the same method without sharing context

Logs

metaPropertiesEnsemble=MetaPropertiesEnsemble(metadataLogDir=Optional.empty, dirs={/var/folders/4g/mxpw3bf903sgt_53jrrx_km40000gp/T/kafka11103251442868018071/broker-0: EMPTY})
io.kroxylicious.testing.kafka.invm.InVMKafkaCluster@4e6f2bb5
org.apache.kafka.clients.admin.KafkaAdminClient@21e20ad5
{virtualClusters:{demo=VirtualCluster[targetCluster=TargetCluster[bootstrapServers='localhost:20273', clientTls=null], clusterNetworkAddressConfigProvider=ClusterNetworkAddressConfigProviderDefinition[type=io.kroxylicious.proxy.internal.clusternetworkaddressconfigprovider.PortPerBrokerClusterNetworkAddressConfigProvider, config=io.kroxylicious.proxy.internal.clusternetworkaddressconfigprovider.PortPerBrokerClusterNetworkAddressConfigProvider$PortPerBrokerClusterNetworkAddressConfigProviderConfig@54657dd2], tls=Optional.empty, logNetwork=false, logFrames=false]},useIoUring:false}
io.kroxylicious.test.tester.DefaultKroxyliciousTester@706eab5d
io.kroxylicious.test.client.KafkaClient@72725ee1
CloseableAdmin[instance=org.apache.kafka.clients.admin.KafkaAdminClient@61f39bb]
metaPropertiesEnsemble=MetaPropertiesEnsemble(metadataLogDir=Optional.empty, dirs={/var/folders/4g/mxpw3bf903sgt_53jrrx_km40000gp/T/kafka14558994476648222760/broker-0: EMPTY})
io.kroxylicious.testing.kafka.invm.InVMKafkaCluster@6724cdec
org.apache.kafka.clients.admin.KafkaAdminClient@6944e53e
{virtualClusters:{demo=VirtualCluster[targetCluster=TargetCluster[bootstrapServers='localhost:16636', clientTls=null], clusterNetworkAddressConfigProvider=ClusterNetworkAddressConfigProviderDefinition[type=io.kroxylicious.proxy.internal.clusternetworkaddressconfigprovider.PortPerBrokerClusterNetworkAddressConfigProvider, config=io.kroxylicious.proxy.internal.clusternetworkaddressconfigprovider.PortPerBrokerClusterNetworkAddressConfigProvider$PortPerBrokerClusterNetworkAddressConfigProviderConfig@52e04737], tls=Optional.empty, logNetwork=false, logFrames=false]},useIoUring:false}
io.kroxylicious.test.tester.DefaultKroxyliciousTester@5bc7e78e
io.kroxylicious.test.client.KafkaClient@1a4d1ab7
CloseableAdmin[instance=org.apache.kafka.clients.admin.KafkaAdminClient@546e61d5]

Additional context
Add any other context about the problem here.

[callaertanthony]

If I set the cluster with @ZooKeeperCluster it works. So it seems that it's the KRaft mode who fails

[k-wall]

I believe you are just seeing the eventually consistent nature of Kafka. Just because you've successfully created a topic does not necessarily mean you'll be able to query it immediately. It may take time to appear. You'd get the same behaviour with just plain old Kafka.

If you add to your test, the problem should go away.

            await().atMost(Duration.ofSeconds(5)).until(() -> admin.listTopics().namesToListings().get(),
                    n -> n.size() == 2);

The @Topic annotation guarantees you that the topic is available.