forked from D-Net/dnet-hadoop
added option to split dumps bigger that 40Gb on different files
This commit is contained in:
parent
78fdb11c3f
commit
14bf2e7238
|
@ -59,7 +59,79 @@ public class MakeTar implements Serializable {
|
||||||
String p_string = p.toString();
|
String p_string = p.toString();
|
||||||
String entity = p_string.substring(p_string.lastIndexOf("/") + 1);
|
String entity = p_string.substring(p_string.lastIndexOf("/") + 1);
|
||||||
|
|
||||||
write(fileSystem, p_string, outputPath + "/" + entity + ".tar", entity);
|
writeMaxSize(fileSystem, p_string, outputPath + "/" + entity, entity, 40);
|
||||||
|
}
|
||||||
|
|
||||||
|
}
|
||||||
|
|
||||||
|
private static TarArchiveOutputStream getTar(FileSystem fileSystem, String outputPath) throws IOException {
|
||||||
|
Path hdfsWritePath = new Path(outputPath);
|
||||||
|
FSDataOutputStream fsDataOutputStream = null;
|
||||||
|
if (fileSystem.exists(hdfsWritePath)) {
|
||||||
|
fileSystem.delete(hdfsWritePath, true);
|
||||||
|
|
||||||
|
}
|
||||||
|
fsDataOutputStream = fileSystem.create(hdfsWritePath);
|
||||||
|
|
||||||
|
return new TarArchiveOutputStream(fsDataOutputStream.getWrappedStream());
|
||||||
|
}
|
||||||
|
|
||||||
|
private static void writeMaxSize(FileSystem fileSystem, String inputPath, String outputPath, String dir_name,
|
||||||
|
int gBperSplit) throws IOException {
|
||||||
|
final long bytesPerSplit = 1024L * 1024L * 1024L * gBperSplit;
|
||||||
|
|
||||||
|
long sourceSize = fileSystem.getContentSummary(new Path(inputPath)).getSpaceConsumed();
|
||||||
|
|
||||||
|
final long numberOfSplits = sourceSize / bytesPerSplit;
|
||||||
|
|
||||||
|
if (numberOfSplits <= 1) {
|
||||||
|
write(fileSystem, inputPath, outputPath + ".tar", dir_name);
|
||||||
|
} else {
|
||||||
|
int partNum = 0;
|
||||||
|
long remainingBytes = sourceSize % bytesPerSplit;
|
||||||
|
|
||||||
|
RemoteIterator<LocatedFileStatus> fileStatusListIterator = fileSystem
|
||||||
|
.listFiles(
|
||||||
|
new Path(inputPath), true);
|
||||||
|
while (sourceSize > 0) {
|
||||||
|
TarArchiveOutputStream ar = getTar(fileSystem, outputPath + "_" + partNum + ".tar");
|
||||||
|
|
||||||
|
long current_size = 0;
|
||||||
|
while (fileStatusListIterator.hasNext() && current_size < bytesPerSplit) {
|
||||||
|
LocatedFileStatus fileStatus = fileStatusListIterator.next();
|
||||||
|
|
||||||
|
Path p = fileStatus.getPath();
|
||||||
|
String p_string = p.toString();
|
||||||
|
if (!p_string.endsWith("_SUCCESS")) {
|
||||||
|
String name = p_string.substring(p_string.lastIndexOf("/") + 1);
|
||||||
|
if (name.trim().equalsIgnoreCase("communities_infrastructures")) {
|
||||||
|
name = "communities_infrastructures.json";
|
||||||
|
}
|
||||||
|
TarArchiveEntry entry = new TarArchiveEntry(dir_name + "/" + name);
|
||||||
|
entry.setSize(fileStatus.getLen());
|
||||||
|
current_size += fileStatus.getLen();
|
||||||
|
ar.putArchiveEntry(entry);
|
||||||
|
|
||||||
|
InputStream is = fileSystem.open(fileStatus.getPath());
|
||||||
|
|
||||||
|
BufferedInputStream bis = new BufferedInputStream(is);
|
||||||
|
|
||||||
|
int count;
|
||||||
|
byte data[] = new byte[1024];
|
||||||
|
while ((count = bis.read(data, 0, data.length)) != -1) {
|
||||||
|
ar.write(data, 0, count);
|
||||||
|
}
|
||||||
|
bis.close();
|
||||||
|
ar.closeArchiveEntry();
|
||||||
|
|
||||||
|
}
|
||||||
|
|
||||||
|
}
|
||||||
|
sourceSize = sourceSize - current_size;
|
||||||
|
partNum += 1;
|
||||||
|
ar.close();
|
||||||
|
}
|
||||||
|
|
||||||
}
|
}
|
||||||
|
|
||||||
}
|
}
|
||||||
|
|
|
@ -0,0 +1,4 @@
|
||||||
|
package eu.dnetlib.dhp.oa.graph.dump.ttl;
|
||||||
|
|
||||||
|
public class OrganizationInfo {
|
||||||
|
}
|
|
@ -0,0 +1,4 @@
|
||||||
|
package eu.dnetlib.dhp.oa.graph.dump.ttl;
|
||||||
|
|
||||||
|
public class Pids {
|
||||||
|
}
|
|
@ -0,0 +1,4 @@
|
||||||
|
package eu.dnetlib.dhp.oa.graph.dump.ttl;
|
||||||
|
|
||||||
|
public class SparkPrepareOrganizationInfo {
|
||||||
|
}
|
Loading…
Reference in New Issue