Restore the explicit closes on solr clients in test to prevent resource starvation

[epugh]

This is a follow up to the PR #3947 where the explicit closing of SolrClient's in DistributedDebugComponentTest failed.

This now passes:

./gradlew :solr:core:test --tests "org.apache.solr.handler.component.DistributedDebugComponentTest" -Ptests.haltonfailure=false "-Ptests.jvmargs=-XX:+UseCompressedOops -XX:+UseParallelGC" -Ptests.multiplier=8

[dsmiley]

It's impossilbe that this fixes the problem. I still see the problem with this change.

[epugh]

Any more details? I was disappointed by this fix, but it did make the Gradle test command pass.

[janhoy]

The test failures I see from DistributedDebugComponentTest in Develocity, e.g. https://develocity.apache.org/s/wtusdzkwyoxow/tests/task/:solr:core:test/details/org.apache.solr.handler.component.DistributedDebugComponentTest/testSimpleSearch?top-execution=1 are all due to SSL being active, and the test stripping https:// prefix in the shard request https://github.com/apache/solr/blob/main/solr/core/src/test/org/apache/solr/handler/component/DistributedDebugComponentTest.java#L72-L73 causing shard handler to fall back to default http:// prefix and failing. No idea why this would be different after switching to TestRule though. Leaving the https:// in the &shards param seems to fix the issue though.

[epugh]

I can try and make the change that you are suggesting @janhoy and see what happens. I swear my change was sorting it out!

[epugh]

Oh, I wonder if somethign about how I was running it locally meant it never did the ssl path?

[hossman]

@RandomizeSSL default odds (when not explicitly configured on the annotation) are increased based on tests.multiplier and tests.nightly