IllegalArgumentException while calling getStreamSegmentInfo.

[Bhupender-Y]

Describe the bug
IllegalArgumentException while calling getStreamSegmentInfo.
To Reproduce
Write the event into stream and parallelly call getStreamSegmentInfo. Run this code in loop, will get this situation

     while (true) {
           streamName = UUID.randomUUID().toString();
           streamManager.createScope("scope");
           streamManager.createStream("scope", streamName, null);
           EventWriterConfig eventWriterConfig = EventWriterConfig.builder()
                                                                  .transactionTimeoutTime(60000)
                                                                  .build();
           EventStreamWriter<String> eventStreamWriter = clientFactory.createEventWriter(streamName, new JavaSerializer<>(), eventWriterConfig);
           new Thread(() -> eventStreamWriter.writeEvent(event).join()).start();
           long length = 0;
           while (length <= 0) {
               length = store.getStreamSegmentInfo(NameUtils.getQualifiedStreamSegmentName("scope", streamName, 0), Duration.ofMinutes(1)).get().getLength();
           }
       }

Screenshots
N/A

Additional information
Add any other information about the problem here, including the root cause if known.

[tkaitchuck]

An example stack trace:

java.lang.IllegalArgumentException: length must be a non-negative number and greater than startOffset.
	at com.google.common.base.Preconditions.checkArgument(Preconditions.java:142)
	at io.pravega.segmentstore.contracts.StreamSegmentInformation.<init>(StreamSegmentInformation.java:78)
	at io.pravega.segmentstore.contracts.StreamSegmentInformation$StreamSegmentInformationBuilder.build(StreamSegmentInformation.java:74)
	at io.pravega.segmentstore.server.containers.StreamSegmentMetadata.getSnapshot(StreamSegmentMetadata.java:377)
	at io.pravega.segmentstore.server.containers.MetadataStore.getSegmentSnapshot(MetadataStore.java:315)
	at io.pravega.segmentstore.server.containers.MetadataStore.getSegmentInfo(MetadataStore.java:284)
	at io.pravega.segmentstore.server.containers.StreamSegmentContainer.getStreamSegmentInfo(StreamSegmentContainer.java:512)
	at io.pravega.segmentstore.server.store.StreamSegmentService.lambda$7(StreamSegmentService.java:119)
	at io.pravega.segmentstore.server.store.SegmentContainerCollection.invoke(SegmentContainerCollection.java:94)
	at io.pravega.segmentstore.server.store.SegmentContainerCollection.invoke(SegmentContainerCollection.java:70)
	at io.pravega.segmentstore.server.store.StreamSegmentService.getStreamSegmentInfo(StreamSegmentService.java:117)

This is also causing build failures:
https://github.com/pravega/pravega/actions/runs/6116044066/job/16600711436

[tkaitchuck]

This is caused by a race condition due to a bug in segment store's design.
In SegmentMetadataUpdateTransaction there is a method apply which applies the relevant updates to the UpdateableSegmentMetadata in this case StreamSegmentMetadata (however the bug exists for all implementations).
In each of these metadata implementations have synchronized methods which update one field at a time, and the apply method updates one field at a time.
The result is that if a read occurs after one field has been updated but before another, the object is in an inconsistent state. Instead the UpdateableSegmentMetadata interface should define a method to update many fields at once with their replacement values. (Or perhaps instead of updating the same object in memory a new object could just be constructed, then the object would be immutable, and it would be up to the SegmentMetadataUpdateTransaction to just replace the object with a new one.

The same bug exists for ContainerMetadataUpdateTransaction