Add maxLifetime to reactive datasource configurations

[kdubb]

This enables configuration of the new maxLifetime value provided by Vert.x 4.4.2 that ensures connections are recycled after a maximum duration.

Because maxLifetime could conceivable be set beyond the maximum value that milliseconds can store in an int (24.8 days) I created a UnitisedTime class. The UnitisedTime class has a unitised method that converts a Duration into the smallest possible TimeUnit that can be stored in an int, starting with milliseconds.

For consistency, I updated the idleTimeout to use the same UnitisedTime.unitised utitlity for conversion.

This was seemingly missed in PR #31873 and is required for use with the Vault Credentials Provider. To this end, I added a blurb about "Time Limited Credentials" to the Credentials Provider docs.

[kdubb]

I can't find where tests are for things like idle-timeout are located. I'm assuming that there should be a place to test the UnitisedTest.unitise method and that configuration properties like idle-timeout and max-lifetime applied to the pool implementations correctly.

[github-actions]

🙈 The PR is closed and the preview is expired.

[tsegismont]

Because maxLifetime could conceivable be set beyond the maximum value that milliseconds can store in an int (24.8 days) I created a UnitisedTime class. The UnitisedTime class has a unitised method that converts a Duration into the smallest possible TimeUnit that can be stored in an int, starting with milliseconds.

This new class is not necessary. In Quarkus, all config properties defined as Duration can take a string value following the standard format: https://docs.oracle.com/javase/8/docs/api/java/time/Duration.html#parse-java.lang.CharSequence-

[tsegismont]

Thank you @kdubb

See my other comment about unitised time

[tsegismont]

Out of curiosity, what's different with the RabbitMQ extension? If the CredentialsProvider has the "expires-at property, couldn't we use the value to configure the pool correctly?

[kdubb]

RabbitMQ connections supports this automatically, it has built support for dynamic credentials.

Vert.x connection pooling does not support this. I made the PR to add max-lifetime to the Vert.x pool and in its current state I believe it would be hard to support this type of feature. Currently the SqlConnectionPool doesn't allow configuration of per-connection properties. If we gain the ability to set properties (specifically something like max lifetime) per connection this could be implemented

[kdubb]

This new class is not necessary. In Quarkus, all config properties defined as Duration can take a string value following the standard format.

@tsegismont I am using that feature, and it is currently used for idle-timeout, but the Vert.x pool takes an int and a TimeUnit for time values; not a duration. Some method of translating between them is necessary.

Previously idle-timeout was converting the duration to long milliseconds and using Math.toIntExact to convert the milliseconds to an int; which throws an ArtithmeticException on overflow. This meant that the max duration it could accept was ~24.8 days, if larger the app would fail with the arithmetic exception.

The unitised utility picks the smallest non-overflowing TimeUnit and uses that. Using the utility allows the configuration duration essentially be unbounded (which is good for max-lifetime). Also, it removes what might be a very "odd" exception on boot if an overflowing time was configured.

[gsmet]

This will need a squash (thinking of the last commit).

Also, I would like to wait for the big Vert.x/datasource/config PR to go in before we rebase and merge this one as this one will conflict with it.

[kdubb]

@gsmet Which PR is that?

[kdubb]

Squashed & rabased. Left the separate doc update as a separate commit.

[gsmet]

@kdubb that would be this small one: #33228

I'll rebase yours once it is in.

[tsegismont]

@tsegismont I am using that feature, and it is currently used for idle-timeout, but the Vert.x pool takes an int and a TimeUnit for time values; not a duration. Some method of translating between them is necessary.

Ok, thanks for the details.

[tsegismont]

Thank you @kdubb

PR rebased.

[gsmet]

@kdubb I rebased the PR on top of my other PR and made the necessary adjustments. I also got rid of the lambdas as we try to avoid lambdas in runtime code.
Let me know if you're still happy with it and we will merge.

[kdubb]

@gsmet LGTM

[quarkus-bot]

Failing Jobs - Building 7503e36

Status	Name	Step	Failures	Logs	Raw logs
✔️	JVM Tests - JDK 11
✖	JVM Tests - JDK 17	Build	Failures	Logs	Raw logs
✔️	JVM Tests - JDK 20

Full information is available in the Build summary check run.

Failures

⚙️ JVM Tests - JDK 17 #

- Failing: integration-tests/kafka-oauth-keycloak 

📦 integration-tests/kafka-oauth-keycloak

✖ io.quarkus.it.kafka.KafkaOauthTest.test - More details - Source on GitHub

java.lang.RuntimeException: java.lang.RuntimeException: Unable to start Quarkus test resource class io.quarkus.it.kafka.KafkaKeycloakTestResource
	at io.quarkus.test.junit.QuarkusTestExtension.throwBootFailureException(QuarkusTestExtension.java:639)
	at io.quarkus.test.junit.QuarkusTestExtension.interceptTestClassConstructor(QuarkusTestExtension.java:711)