When Using Liquibase in Native Mode a ServiceConfigurationError is Logged

[gcw-it]

Describe the bug

When Using Liquibase in Native Mode a ServiceConfigurationError is Logged.
This happens at least with Postgresql and H2.
The issue occurs at least since Quarkus 3.8.2
The error occurs for a local as well as an in-container build.
There seem to be no negative consequences of this error, besides polluting the application log.

Expected behavior

No Error is logged.

Actual behavior

At Quarkus start the following error is logged:

INFO  [liq.servicelocator] (main) Cannot load service: java.util.ServiceConfigurationError: liquibase.command.CommandStep: Provider liquibase.command.core.StartH2CommandStep not found
	at java.base@21.0.3/java.util.ServiceLoader.fail(ServiceLoader.java:593)
	at java.base@21.0.3/java.util.ServiceLoader$LazyClassPathLookupIterator.nextProviderClass(ServiceLoader.java:1219)
	at java.base@21.0.3/java.util.ServiceLoader$LazyClassPathLookupIterator.hasNextService(ServiceLoader.java:1228)
	at java.base@21.0.3/java.util.ServiceLoader$LazyClassPathLookupIterator.hasNext(ServiceLoader.java:1273)
	at java.base@21.0.3/java.util.ServiceLoader$2.hasNext(ServiceLoader.java:1309)
	at java.base@21.0.3/java.util.ServiceLoader$3.hasNext(ServiceLoader.java:1393)
	at liquibase.servicelocator.StandardServiceLocator.findInstances(StandardServiceLocator.java:22)
	at liquibase.command.CommandFactory.findAllInstances(CommandFactory.java:244)
	at liquibase.command.CommandFactory.computePipelineForCommandDefinition(CommandFactory.java:71)
	at liquibase.command.CommandFactory.getCommandDefinition(CommandFactory.java:50)
	at liquibase.command.CommandScope.<init>(CommandScope.java:64)
	at liquibase.Liquibase.lambda$validate$25(Liquibase.java:1365)
	at liquibase.Scope.lambda$child$0(Scope.java:190)
	at liquibase.Scope.child(Scope.java:199)
	at liquibase.Scopefor child(Scope.java:189)
	at liquibase.Scope.child(Scope.java:168)
	at liquibase.Liquibase.runInScope(Liquibase.java:1436)
	at liquibase.Liquibase.validate(Liquibase.java:1364)
	at io.quarkus.liquibase.runtime.LiquibaseRecorder.doStartActions(LiquibaseRecorder.java:79)
	at io.quarkus.deployment.steps.LiquibaseProcessor$startLiquibase1744275855.deploy_0(Unknown Source)
	at io.quarkus.deployment.steps.LiquibaseProcessor$startLiquibase1744275855.deploy(Unknown Source)
	at io.quarkus.runner.ApplicationImpl.doStart(Unknown Source)
	at io.quarkus.runtime.Application.start(Application.java:101)
	at io.quarkus.runtime.ApplicationLifecycleManager.run(ApplicationLifecycleManager.java:111)
	at io.quarkus.runtime.Quarkus.run(Quarkus.java:71)
	at io.quarkus.runtime.Quarkus.run(Quarkus.java:44)
	at io.quarkus.runtime.Quarkus.run(Quarkus.java:124)
	at io.quarkus.runner.GeneratedMain.main(Unknown Source)

The error is exactly the same for Postgresql and H2.
It probably occurs for other databases as well.

How to Reproduce?

1. Use the liquibase-quickstart from quarkus-quickstarters.
2. Run a H2 instance with remote database creation enabled.
3. Compile and run the app in native mode.

Output of uname -a or ver

Darwin 23.4.0 Darwin Kernel Version 23.4.0: Fri Mar 15 00:11:05 PDT 2024; root:xnu-10063.101.17~1/RELEASE_X86_64 x86_64

Output of java -version

java 21.0.3 2024-04-16 LTS Java(TM) SE Runtime Environment Oracle GraalVM 21.0.3+7.1 (build 21.0.3+7-LTS-jvmci-23.1-b37) Java HotSpot(TM) 64-Bit Server VM Oracle GraalVM 21.0.3+7.1 (build 21.0.3+7-LTS-jvmci-23.1-b37, mixed mode, sharing)

Mandrel or GraalVM version (if different from Java)

No response

Quarkus version or git rev

3.10.0

Build tool (ie. output of mvnw --version or gradlew --version)

Apache Maven 3.9.6 (bc0240f3c744dd6b6ec2920b3cd08dcc295161ae)

Additional information

No response

[quarkus-bot]

/cc @Karm (mandrel), @andrejpetras (liquibase), @galderz (mandrel), @geoand (liquibase), @gsmet (liquibase), @radcortez (config), @zakkak (mandrel,native-image)

[geoand]

cc @famod you looked into this last

[HerrDerb]

Does this fix only the h2 issue? We too experience this with postgres

[geoand]

Right, this was closed by mistake.

[geoand]

How does one reproduce this for postgres?

[HerrDerb]

I'll try to provide a reproducer, my first guess is the same just with a postgres driver

[gcw-it]

You can use the steps outlined in issue 40574 under How to Reproduce? to replicate the error for postgres.

[HerrDerb]

You can use the steps outlined in issue 40574 under How to Reproduce? to replicate the error for postgres.

Brilliant, thank you.
I can approve that we also use quarkus.liquibase.migrate-at-start=true
We seem NOT to set quarkus.liquibase.clean-at-start=true

[geoand]

I can confirm that I get:

liquibase.exception.DatabaseException: Error executing SQL UPDATE public.databasechangeloglock SET LOCKED = FALSE, LOCKEDBY = NULL, LOCKGRANTED = NULL WHERE ID = 1: ERROR: relation "public.databasechangeloglock" does not exist

when using Postgres, but I have no idea what that means.

[geoand]

Nevermind the comment above - I see why that happens in native tests.
So the real questions becomes why does

Cannot load service: java.util.ServiceConfigurationError: liquibase.command.CommandStep: Provider liquibase.command.core.StartH2CommandStep not found

show up in the logs

[HerrDerb]

Yes, especially when you don't have an h2 dependency.

[geoand]

My best guess is that we should always do () -> true from

quarkus/extensions/liquibase/deployment/src/main/java/io/quarkus/liquibase/deployment/LiquibaseProcessor.java

Line 256 in 509ec82

private static Predicate<String> commandStepPredicate(Capabilities capabilities) {

.

Do you want to try that out in your applications?

[gcw-it]

Brilliant, thank you. I can approve that we also use quarkus.liquibase.migrate-at-start=true We seem NOT to set quarkus.liquibase.clean-at-start=true

You're right. This setting is only required to reproduce the issue addressed in the linked topic.
In this context the setting is unnecessary.