HDDS-15623. Disk check may not completely read test file

[adoroszlai]

What changes were proposed in this pull request?

1. Disk read/write check may not completely read the test file with InputStream.read().
   

   ozone/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/utils/DiskCheckUtil.java

   Lines 167 to 175 in 9649e0e

   	// Read data back from the test file.
   	byte[] readBytes = new byte[numBytesToWrite];
   	try (InputStream fis = Files.newInputStream(testFile.toPath())) {
   	int numBytesRead = fis.read(readBytes);
   	if (numBytesRead != numBytesToWrite) {
   	logError(storageDir, String.format("%d bytes written to file %s " +
   	"but %d bytes were read back.", numBytesToWrite,
   	testFile.getAbsolutePath(), numBytesRead));
   	return false;

   
   Use IOUtils.read to fully read the input.
2. Delete the test file even if the check fails for any reason.
3. Convert File.toPath once.

https://issues.apache.org/jira/browse/HDDS-15623

How was this patch tested?

Updated unit test.

CI:
https://github.com/adoroszlai/ozone/actions/runs/27831059411

[ashishkumar50]

@adoroszlai Thanks for the patch.

[ashishkumar50]

It would be better to delete in finally, instead in each catch exception.

[adoroszlai]

The file is deleted as part of the check in normal case:

ozone/hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/common/utils/DiskCheckUtil.java

Lines 205 to 213 in f27e18d

	// Delete the file.
	if (!testFile.delete()) {
	logError(storageDir, String.format("Could not delete file %s " +
	"for volume check.", testFile.getAbsolutePath()));
	return false;
	}
	// If all checks passed, the volume is healthy.
	return true;

So we cannot delete in finally of read check.

[ashishkumar50]

Makes sense, in normal scenario delete is verified.

[ashishkumar50]

@adoroszlai thanks for the clarification, patch LGTM.