Create Tar archive from directory on S3 using AWS Lambda

Question

I need to extract a bunch of zip files stored on s3 and add them to a tar archive and store that archive on s3. it is likely that that the sum of the zip files will greater than the 512mb local storage allowed from lambda functions. I have a partial souldtion that gets the objects from s3 extracts them and puts them in a s3 object without using the lambda local storage.

Extract object Thread

public class ExtractObject implements Runnable{

    private String objectName;
    private String uuid;
    private final byte[] buffer = new byte[1024];

    public ExtractAdvert(String name, String uuid) {
        this.objectName= name;
        this.uuid= uuid;
    }

    @Override
    public void run() {
        final String srcBucket = "my-bucket-name";
        final AmazonS3 s3Client = new AmazonS3Client();

        try {
            S3Object s3Object = s3Client.getObject(new GetObjectRequest(srcBucket, objectName));
            ZipInputStream zis = new ZipInputStream(s3Object.getObjectContent());
            ZipEntry entry = zis.getNextEntry();

            while(entry != null) {
                String fileName = entry.getName();
                String mimeType = FileMimeType.fromExtension(FilenameUtils.getExtension(fileName)).mimeType();
                System.out.println("Extracting " + fileName + ", compressed: " + entry.getCompressedSize() + " bytes, extracted: " + entry.getSize() + " bytes, mimetype: " + mimeType);
                ByteArrayOutputStream outputStream = new ByteArrayOutputStream();
                int len;
                while ((len = zis.read(buffer)) > 0) {
                    outputStream.write(buffer, 0, len);
                }
                InputStream is = new ByteArrayInputStream(outputStream.toByteArray());
                ObjectMetadata meta = new ObjectMetadata();
                meta.setContentLength(outputStream.size());
                meta.setContentType(mimeType);
                System.out.println("##### " + srcBucket + ", " + FilenameUtils.getFullPath(objectName) + "tmp" + File.separator + uuid + File.separator + fileName);

                // Add this to tar archive instead of putting back to s3
                s3Client.putObject(srcBucket, FilenameUtils.getFullPath(objectName) + "tmp" + File.separator + uuid + File.separator + fileName, is, meta);
                is.close();
                outputStream.close();
                entry = zis.getNextEntry();
            }
            zis.closeEntry();
            zis.close();
        } catch (IOException ioe) {
                System.out.println(ioe.getMessage());
        }
    }
}

---

this runs for each object that needs to be extracted and saves them in a s3 object in the structure needed for the tar file.

I think what i need is instead of putting the object back to s3 is to keep it in memory and add it to a tar archive. and upload that but after a lot of looking around and trial and error i have not created a successful tar file. The main issue is i can't use the tmp directory in lambda.

---

Edit should i be creating the tar file as i go instead of putting objects to s3? (see comment // Add this to tar archive instead of putting back to s3) if so how do i create a tar stream without a storing it locally?

---

EDIT 2: Attempt at taring the files

ListObjectsV2Request req = new ListObjectsV2Request().withBucketName(bucketName);
ListObjectsV2Result result;

ByteArrayOutputStream baos = new ByteArrayOutputStream();
TarArchiveOutputStream tarOut = new TarArchiveOutputStream(baos);

do {
    result = s3Client.listObjectsV2(req);

    for (S3ObjectSummary objectSummary : result.getObjectSummaries()) {

        if(objectSummary.getKey().startsWith("tmp/") )  {
            System.out.printf(" - %s (size: %d)\n", objectSummary.getKey(), objectSummary.getSize());
            S3Object s3Object = s3Client.getObject(new GetObjectRequest(bucketName, objectSummary.getKey()));
            InputStream is = s3Object.getObjectContent(); 
            System.out.println("Pre Create entry");
            TarArchiveEntry archiveEntry = new TarArchiveEntry(IOUtils.toByteArray(is));
            // Getting following exception above
            // IllegalArgumentException: Invalid byte 111 at offset 7 in ' positio' len=8
            System.out.println("Pre put entry");
            tarOut.putArchiveEntry(archiveEntry);
            System.out.println("Post put entry");
        }
    }

    String token = result.getNextContinuationToken();
    System.out.println("Next Continuation Token: " + token);
    req.setContinuationToken(token);
} while (result.isTruncated());

ObjectMetadata metadata = new ObjectMetadata();
InputStream is = new ByteArrayInputStream(baos.toByteArray());
s3Client.putObject(new PutObjectRequest(bucketName, bucketFolder + "tar-file", is, metadata));

Answer 1

I have found a solution to this and it very similar to my attempt in Edit 2 above.

private final String bucketName = "bucket-name";
private final String bucketFolder = "tmp/";
private final String tarKey = "tar-dir/tared-file.tar";

private void createTar() throws IOException, ArchiveException {
    ListObjectsV2Request req = new ListObjectsV2Request().withBucketName(bucketName);
    ListObjectsV2Result result;

    ByteArrayOutputStream baos = new ByteArrayOutputStream();
    TarArchiveOutputStream tarOut = new TarArchiveOutputStream(baos);

    do {
        result = s3Client.listObjectsV2(req);

        for (S3ObjectSummary objectSummary : result.getObjectSummaries()) {
            if (objectSummary.getKey().startsWith(bucketFolder)) {
                S3Object s3Object = s3Client.getObject(new GetObjectRequest(bucketName, objectSummary.getKey()));
                InputStream is = s3Object.getObjectContent();

                String s3Key = objectSummary.getKey();
                String tarPath = s3Key.substring(s3Key.indexOf('/') + 1, s3Key.length());
                s3Key.lastIndexOf('.'));

                byte[] ba = IOUtils.toByteArray(is);

                TarArchiveEntry archiveEntry = new TarArchiveEntry(tarPath);
                archiveEntry.setSize(ba.length);
                tarOut.putArchiveEntry(archiveEntry);
                tarOut.write(ba);
                tarOut.closeArchiveEntry();
            }
        }

        String token = result.getNextContinuationToken();
        System.out.println("Next Continuation Token: " + token);
        req.setContinuationToken(token);
    } while (result.isTruncated());

    ObjectMetadata metadata = new ObjectMetadata();
    InputStream is = baos.toInputStream();
    metadata.setContentLength(baos.size());
    s3Client.putObject(new PutObjectRequest(bucketName, tarKey, is, metadata));
}