Fixed integration tests

[dlg99]

Motivation

KafkaApiTest is failing

Modifications

1. conversion between Pulsar topic name and Kafka TopicPartition ended up with TopicPartition using name with "-partition-"

2. Seek was not working correctly:

PulsarKafkaConsumer seeks to beginning, as asked.
Clears lastReceivedOffset in the process.

on poll it checks

            if (lastReceivedOffset.get(tp) == null && !unpolledPartitions.contains(tp)) {
                	log.info("When polling offsets, invalid offsets were detected. Resetting topic partition {}", tp);
                	resetOffsets(tp);
            } 

seek didn't update unpolledPartitions - reset offset uses default strategy to reset => seeks to the end

Verifying this change

• Make sure that the change passes the CI checks.

(Please pick either of the following options)

This change is already covered by existing tests, such as KafkaApiTest.

Does this pull request potentially affect one of the following parts:

If yes was chosen, please highlight the changes

No

Documentation

• Does this pull request introduce a new feature? NO

[dlg99]

KafkaProducerSimpleConsumerTest dosn't fail on local runs but failed on CI. I added couple of delays, it worked.

KafkaApisTest passes now.

Failing integration test is org.apache.pulsar.spark.SparkStreamingPulsarReceiverTest, can be addressed separately

[lhotari]

LGTM

[dlg99]

also fixed SparkStreamingPulsarReceiverTest

[lhotari]

@codelipenghui @sijie Please review. This PR makes CI green and we should then be ready for the 2.8.0 release of pulsar-adapters.

[codelipenghui]

Looks good to me, just left a minor comment about the test.

[codelipenghui]

It's better to avoid using sleep in the test, could you please use Awaitility instead?

[dlg99]

I agree that it would be nice, but I suggest we merge this to unblock the release + add an issue to fix it later (it will take time to figure out what to wait for specifically, especially given that this test failed only on CI and passed on local runs) + there are other tests that rely on sleep: https://github.com/apache/pulsar-adapters/search?q=Thread.Sleep

[eolivelli]

I am fine for keeping Sleep for now.
it is not clear what we have to wait for at the moment

[eolivelli]

I am fine with the current approach, especially if @dlg99 will follow up and create an issue to remove the Thread.sleep

[eolivelli]

I am fine for keeping Sleep for now.
it is not clear what we have to wait for at the moment

[eolivelli]

@codelipenghui
I like the fact that with this patch all tests passed