HDDS-14219. Add metrics for ReadBlock

[rich7420]

What changes were proposed in this pull request?

This PR adds metrics tracking for readBlock operations and unit test.

What is the link to the Apache JIRA

HDDS-14219

How was this patch tested?

test all passed locally.
https://github.com/rich7420/ozone/actions/runs/20390167387/

[jojochuang]

must call close() on this object to release file description upon close.

RandomAccessFileChannel should have been made a Closeable object. Too bad that's not the case.

[jojochuang]

https://issues.apache.org/jira/browse/HDDS-14370

[rich7420]

Ok,get it!
thanks for the review!

[jojochuang]

I thought I added this comment too but it didn't show up before. Sorry for back and forth.

[jojochuang]

this is an expensive operation because it involves reading from rocksdb. And for the most part this blockData is not going to be used since readBlock usually has length. So this is a waste.

IMO, we should let readBlockImpl() to return the length it actually read and use that as the counter for bytesRead.

[jojochuang]

request change not approval.

[rich7420]

I've tried to resolve this problem and update.

[ss77892]

We should also update the Volume statistic. Something like:

  final long startTime = Time.monotonicNow();
  final long bytesRead = readBlockImpl(request, blockFile, kvContainer, streamObserver, false);
  long endTime = Time.monotonicNow();
  KeyValueContainerData containerData = (KeyValueContainerData) kvContainer
      .getContainerData();
  HddsVolume volume = containerData.getVolume();
  if (volume != null) {
    volume.getVolumeIOStats().incReadTime(endTime - startTime);
    volume.getVolumeIOStats().incReadOpCount();
    volume.getVolumeIOStats().incReadBytes(bytesRead);
  }

[jojochuang]

@ss77892 IMO the volume level metrics should be updated within RandomAccessFileChannel itself.

But yeah, since this request readBlock does not use the regular ChunkUtils/FilePerBlockStrategy, volume level metrics are not automatically updated.