ssize_t HdfsFile::pread(void *ptr, off64_t offset, size_t count)
{
boost::mutex::scoped_lock lock(m_mutex);
tSize ret = hdfsPread(m_fs, m_file, offset, ptr, count);
// this is an observed case when trying to read from an open file handle
// that is now stale (because of rewrite/whatever). If we see this,
// assume that the file has likely been rewritten and a subsequent attempt
// to read will be successful if we reopen the file.
tSize retryCount = 0;
while (ret != (tSize) count)
{
//if (ret != count)
// IDBLogger::logRW("pread_retry", m_fname, this, offset, count, ret);
if (retryCount >= 50)
break;
usleep(1000);
reopen();
ret = hdfsPread(m_fs, m_file, offset, ptr, count);
retryCount++;
}
if( IDBLogger::isEnabled() )
IDBLogger::logRW("pread", m_fname, this, offset, count, ret);
return ret;
}
qioerr hdfs_preadv (void* file, const struct iovec *vector, int count, off_t offset, ssize_t* num_read_out, void* fs)
{
ssize_t got;
ssize_t got_total;
qioerr err_out = 0;
int i;
STARTING_SLOW_SYSCALL;
err_out = 0;
got_total = 0;
for(i = 0; i < count; i++) {
got = hdfsPread(to_hdfs_fs(fs)->hfs, to_hdfs_file(file)->file, offset + got_total, (void*)vector[i].iov_base, vector[i].iov_len);
if( got != -1 ) {
got_total += got;
} else {
err_out = qio_mkerror_errno();
break;
}
if(got != (ssize_t)vector[i].iov_len ) {
break;
}
}
if( err_out == 0 && got_total == 0 && sys_iov_total_bytes(vector, count) != 0 )
err_out = qio_int_to_err(EEOF);
*num_read_out = got_total;
DONE_SLOW_SYSCALL;
return err_out;
}
int BlockManager::loadFromHdfs(const ChunkID& chunk_id, void* const &desc,const unsigned & length){
lock.acquire();
int ret;
int offset=chunk_id.chunk_off;
hdfsFS fs=hdfsConnect(Config::hdfs_master_ip.c_str(),Config::hdfs_master_port);
hdfsFile readFile=hdfsOpenFile(fs,chunk_id.partition_id.getPathAndName().c_str(),O_RDONLY,0,0,0);
hdfsFileInfo *hdfsfile=hdfsGetPathInfo(fs,chunk_id.partition_id.getPathAndName().c_str());// to be refined after communicating with Zhang Lei
if(!readFile){
logging_->elog("Fail to open file [%s].Reason:%s",chunk_id.partition_id.getPathAndName().c_str(),strerror(errno));
hdfsDisconnect(fs);
lock.release();
return -1;
}
else{
logging_->log("file [%s] is opened for offset[%d]\n",chunk_id.partition_id.getPathAndName().c_str(),offset);
}
long int start_pos=CHUNK_SIZE*offset;
if(start_pos<hdfsfile->mSize){
ret=hdfsPread(fs,readFile,start_pos,desc,length);
}else{
lock.release();
ret= -1;
}
hdfsCloseFile(fs,readFile);
hdfsDisconnect(fs);
lock.release();
return ret;
}
// 这个函数返回一个构造好的chunkid和每个chunk的指针
// 这个里面的chunkId肯定是要在blockManager注册然后汇报信息的
// put的话也是会这样的，可以将这个函数中调用put然后统一汇报信息的接口
ChunkInfo BlockManager::loadFromHdfs(string file_name){
// 由此函数得到的<blockId,指针>
ChunkInfo ci;
string file_name_former,file_name_latter;
unsigned pos=file_name.rfind("$");
file_name_former=file_name.substr(0,pos);
file_name_latter=file_name.substr(pos+1,file_name.length());
int offset=atoi(file_name_latter.c_str());
hdfsFS fs=hdfsConnect(Config::hdfs_master_ip.c_str(),Config::hdfs_master_port);
hdfsFile readFile=hdfsOpenFile(fs,file_name_former.c_str(),O_RDONLY,0,0,0);
hdfsFileInfo *hdfsfile=hdfsGetPathInfo(fs,file_name_former.c_str());
if(!readFile){
cout<<"open file error"<<endl;
}
unsigned length=0;
length=length+CHUNK_SIZE*offset;
if(length<hdfsfile->mSize){
void *rt=malloc(CHUNK_SIZE); //newmalloc
tSize bytes_num=hdfsPread(fs,readFile,length,rt,CHUNK_SIZE);
ostringstream chunkid;
chunkid<<file_name.c_str()<<"$"<<offset;
// ci.chunkId=chunkid.gestr().c_str();
ci.hook=rt;
}else{
ostringstream chunkid;
chunkid<<file_name.c_str()<<"$"<<offset;
// ci.chunkId=chunkid.str().c_str();
ci.hook=0;
}
hdfsCloseFile(fs,readFile);
hdfsDisconnect(fs);
return ci;
}
int hdfs_file_pread(struct back_storage *storage, bs_file_t file, \
const char *read_buff, uint32_t read_len, uint64_t pos){
//HLOG_DEBUG("hdfs -- enter func %s", __func__);
hdfsFile hfile = (hdfsFile)file;
//HLOG_DEBUG("hdfs -- leave func %s", __func__);
return hdfsPread((hdfsFS)storage->fs_handler, \
hfile, pos, (void *) read_buff, read_len);
}
/**
* hdfsRead - Read data from an open file.
*
* @param fs The configured filesystem handle.
* @param file The file handle.
* @param buffer The buffer to copy read bytes into.
* @param length The length of the buffer.
* @return Returns the number of bytes actually read, possibly less
* than than length; -1 on error.
*/
tSize
hdfsRead(hdfsFS fs, hdfsFile file, void* buffer, tSize length)
{
struct hdfsFile_internal *f = file;
tSize res;
res = hdfsPread(fs, file, f->fi_offset, buffer, length);
if (res != -1)
f->fi_offset += length;
return res;
}
bool HDFSChunkReaderIterator::nextBlock(BlockStreamBase*& block){
if(cur_block_>=number_of_blocks_){
lock_.acquire();
return false;
}
tSize bytes_num=hdfsPread(fs_,hdfs_fd_,cur_block_*block_size_,block_buffer_->getBlock(),CHUNK_SIZE);
if(bytes_num==block_size_){
printf("cur block=%d\n",cur_block_);
cur_block_++;
block->constructFromBlock(*block_buffer_);
lock_.release();
return true;
}
else{
cur_block_++;
lock_.release();
return false;
}
}
ssize_t HdfsFile::real_pread(void *ptr, off64_t offset, size_t count)
{
int savedErrno = -1;
size_t progress = 0;
uint8_t *ptr8 = (uint8_t *) ptr;
tSize ret = 0;
// this is an observed case when trying to read from an open file handle
// that is now stale (because of rewrite/whatever). If we see this,
// assume that the file has likely been rewritten and a subsequent attempt
// to read will be successful if we reopen the file.
/* Revised s.t. progress is not counted against the number of retry
attempts. */
tSize retryCount = 0;
while (progress < count && retryCount < 100)
{
ret = hdfsPread(m_fs, m_file, offset + progress, &ptr8[progress], count - progress);
savedErrno = errno;
if (ret > 0) { // making progress, reset the retry counter
retryCount = 0;
progress += ret;
}
else { // an error or EOF
// cout << "retrying... progress = " << progress << " count = " << count <<
// " filename = " << name() << endl;
if (retryCount < 10)
usleep(1000);
else
usleep(200000);
reopen(); // keep retrying, regardless of an error on reopen
retryCount++;
}
}
if( IDBLogger::isEnabled() )
IDBLogger::logRW("pread", m_fname, this, offset, count, ret);
errno = savedErrno;
return progress;
}
int main(int argc, char **argv) {
hdfsFS fs = hdfsConnect("default", 0);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs!\n");
exit(-1);
}
hdfsFS lfs = hdfsConnect(NULL, 0);
if(!lfs) {
fprintf(stderr, "Oops! Failed to connect to 'local' hdfs!\n");
exit(-1);
}
const char* writePath = "/tmp/testfile.txt";
{
//Write tests
hdfsFile writeFile = hdfsOpenFile(fs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", writePath);
char* buffer = "Hello, World!";
tSize num_written_bytes = hdfsWrite(fs, writeFile, (void*)buffer, strlen(buffer)+1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
tOffset currentPos = -1;
if ((currentPos = hdfsTell(fs, writeFile)) == -1) {
fprintf(stderr,
"Failed to get current file position correctly! Got %ld!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %ld\n", currentPos);
if (hdfsFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", writePath);
hdfsCloseFile(fs, writeFile);
}
{
//Read tests
const char* readPath = "/tmp/testfile.txt";
int exists = hdfsExists(fs, readPath);
if (exists) {
fprintf(stderr, "Failed to validate existence of %s\n", readPath);
exit(-1);
}
hdfsFile readFile = hdfsOpenFile(fs, readPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", readPath);
exit(-1);
}
fprintf(stderr, "hdfsAvailable: %d\n", hdfsAvailable(fs, readFile));
tOffset seekPos = 1;
if(hdfsSeek(fs, readFile, seekPos)) {
fprintf(stderr, "Failed to seek %s for reading!\n", readPath);
exit(-1);
}
tOffset currentPos = -1;
if((currentPos = hdfsTell(fs, readFile)) != seekPos) {
fprintf(stderr,
"Failed to get current file position correctly! Got %ld!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %ld\n", currentPos);
static char buffer[32];
tSize num_read_bytes = hdfsRead(fs, readFile, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
num_read_bytes = hdfsPread(fs, readFile, 0, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
hdfsCloseFile(fs, readFile);
}
int totalResult = 0;
int result = 0;
{
//Generic file-system operations
const char* srcPath = "/tmp/testfile.txt";
const char* dstPath = "/tmp/testfile2.txt";
fprintf(stderr, "hdfsCopy(remote-local): %s\n", ((result = hdfsCopy(fs, srcPath, lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(local-local): %s\n", ((result = hdfsMove(lfs, srcPath, lfs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(remote-local): %s\n", ((result = hdfsMove(fs, srcPath, lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsRename: %s\n", ((result = hdfsRename(fs, dstPath, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
const char* slashTmp = "/tmp";
const char* newDirectory = "/tmp/newdir";
fprintf(stderr, "hdfsCreateDirectory: %s\n", ((result = hdfsCreateDirectory(fs, newDirectory)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsSetReplication: %s\n", ((result = hdfsSetReplication(fs, srcPath, 2)) ? "Failed!" : "Success!"));
totalResult += result;
char buffer[256];
const char *resp;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ? buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsSetWorkingDirectory: %s\n", ((result = hdfsSetWorkingDirectory(fs, slashTmp)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ? buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsGetDefaultBlockSize: %ld\n", hdfsGetDefaultBlockSize(fs));
fprintf(stderr, "hdfsGetCapacity: %ld\n", hdfsGetCapacity(fs));
fprintf(stderr, "hdfsGetUsed: %ld\n", hdfsGetUsed(fs));
hdfsFileInfo *fileInfo = NULL;
if((fileInfo = hdfsGetPathInfo(fs, slashTmp)) != NULL) {
fprintf(stderr, "hdfsGetPathInfo - SUCCESS!\n");
fprintf(stderr, "Name: %s, ", fileInfo->mName);
fprintf(stderr, "Type: %c, ", (char)(fileInfo->mKind));
fprintf(stderr, "Replication: %d, ", fileInfo->mReplication);
fprintf(stderr, "BlockSize: %ld, ", fileInfo->mBlockSize);
fprintf(stderr, "Size: %ld, ", fileInfo->mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileInfo->mLastMod));
fprintf(stderr, "Owner: %s, ", fileInfo->mOwner);
fprintf(stderr, "Group: %s, ", fileInfo->mGroup);
char permissions[10];
permission_disp(fileInfo->mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileInfo->mPermissions, permissions);
hdfsFreeFileInfo(fileInfo, 1);
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetPathInfo for %s - FAILED!\n", slashTmp);
}
hdfsFileInfo *fileList = 0;
int numEntries = 0;
if((fileList = hdfsListDirectory(fs, slashTmp, &numEntries)) != NULL) {
int i = 0;
for(i=0; i < numEntries; ++i) {
fprintf(stderr, "Name: %s, ", fileList[i].mName);
fprintf(stderr, "Type: %c, ", (char)fileList[i].mKind);
fprintf(stderr, "Replication: %d, ", fileList[i].mReplication);
fprintf(stderr, "BlockSize: %ld, ", fileList[i].mBlockSize);
fprintf(stderr, "Size: %ld, ", fileList[i].mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileList[i].mLastMod));
fprintf(stderr, "Owner: %s, ", fileList[i].mOwner);
fprintf(stderr, "Group: %s, ", fileList[i].mGroup);
char permissions[10];
permission_disp(fileList[i].mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileList[i].mPermissions, permissions);
}
hdfsFreeFileInfo(fileList, numEntries);
} else {
if (errno) {
totalResult++;
fprintf(stderr, "waah! hdfsListDirectory - FAILED!\n");
} else {
fprintf(stderr, "Empty directory!\n");
}
}
char*** hosts = hdfsGetHosts(fs, srcPath, 0, 1);
if(hosts) {
fprintf(stderr, "hdfsGetHosts - SUCCESS! ... \n");
int i=0;
while(hosts[i]) {
int j = 0;
while(hosts[i][j]) {
fprintf(stderr,
"\thosts[%d][%d] - %s\n", i, j, hosts[i][j]);
++j;
}
++i;
}
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetHosts - FAILED!\n");
}
char *newOwner = "root";
// setting tmp dir to 777 so later when connectAsUser nobody, we can write to it
short newPerm = 0666;
// chown write
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, NULL, "users")) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, newOwner, NULL)) ? "Failed!" : "Success!"));
totalResult += result;
// chmod write
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, writePath, newPerm)) ? "Failed!" : "Success!"));
totalResult += result;
sleep(2);
tTime newMtime = time(NULL);
tTime newAtime = time(NULL);
// utime write
fprintf(stderr, "hdfsUtime: %s\n", ((result = hdfsUtime(fs, writePath, newMtime, newAtime)) ? "Failed!" : "Success!"));
totalResult += result;
// chown/chmod/utime read
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, writePath);
fprintf(stderr, "hdfsChown read: %s\n", ((result = (strcmp(finfo->mOwner, newOwner) != 0)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChmod read: %s\n", ((result = (finfo->mPermissions != newPerm)) ? "Failed!" : "Success!"));
totalResult += result;
// will later use /tmp/ as a different user so enable it
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, "/tmp/", 0777)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr,"newMTime=%ld\n",newMtime);
fprintf(stderr,"curMTime=%ld\n",finfo->mLastMod);
fprintf(stderr, "hdfsUtime read (mtime): %s\n", ((result = (finfo->mLastMod != newMtime)) ? "Failed!" : "Success!"));
totalResult += result;
// No easy way to turn on access times from hdfs_test right now
// fprintf(stderr, "hdfsUtime read (atime): %s\n", ((result = (finfo->mLastAccess != newAtime)) ? "Failed!" : "Success!"));
// totalResult += result;
hdfsFreeFileInfo(finfo, 1);
// Clean up
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, newDirectory)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsExists: %s\n", ((result = hdfsExists(fs, newDirectory)) ? "Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
}
totalResult += (hdfsDisconnect(fs) != 0);
{
//
// Now test as connecting as a specific user
// This is only meant to test that we connected as that user, not to test
// the actual fs user capabilities. Thus just create a file and read
// the owner is correct.
const char *tuser = "******";
const char* writePath = "/tmp/usertestfile.txt";
const char **groups = (const char**)malloc(sizeof(char*)* 2);
groups[0] = "users";
groups[1] = "nobody";
fs = hdfsConnectAsUser("default", 0, tuser, groups, 2);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs as user %s!\n",tuser);
exit(-1);
}
hdfsFile writeFile = hdfsOpenFile(fs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", writePath);
char* buffer = "Hello, World!";
tSize num_written_bytes = hdfsWrite(fs, writeFile, (void*)buffer, strlen(buffer)+1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
if (hdfsFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", writePath);
hdfsCloseFile(fs, writeFile);
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, writePath);
fprintf(stderr, "hdfs new file user is correct: %s\n", ((result = (strcmp(finfo->mOwner, tuser) != 0)) ? "Failed!" : "Success!"));
totalResult += result;
}
totalResult += (hdfsDisconnect(fs) != 0);
if (totalResult != 0) {
return -1;
} else {
return 0;
}
}
NABoolean HHDFSFileStats::populate(hdfsFS fs, hdfsFileInfo *fileInfo,
Int32& samples,
NABoolean doEstimation, char recordTerminator,
NABoolean isSequenceFile)
{
NABoolean result = TRUE;
// copy fields from fileInfo
fileName_ = fileInfo->mName;
replication_ = (Int32) fileInfo->mReplication;
totalSize_ = (Int64) fileInfo->mSize;
blockSize_ = (Int64) fileInfo->mBlockSize;
modificationTS_ = fileInfo->mLastMod;
numFiles_ = 1;
isSequenceFile_ = isSequenceFile;
Int64 sampleBufferSize = MINOF(blockSize_, 65536);
NABoolean sortHosts = (CmpCommon::getDefault(HIVE_SORT_HDFS_HOSTS) == DF_ON);
sampleBufferSize = MINOF(sampleBufferSize,totalSize_/10);
if (doEstimation && sampleBufferSize > 100) {
//
// Open the hdfs file to estimate record length. Read one block at
// a time searching for <s> instances of record separators. Stop reading
// when either <s> instances have been found or a partial number of
// instances have and we have exhausted all data content in the block.
// We will keep reading if the current block does not contain
// any instance of the record separator.
//
hdfsFile file =
hdfsOpenFile(fs, fileInfo->mName,
O_RDONLY,
sampleBufferSize, // buffer size
0, // replication, take the default size
fileInfo->mBlockSize // blocksize
);
if ( file != NULL ) {
tOffset offset = 0;
tSize bufLen = sampleBufferSize;
char* buffer = new (heap_) char[bufLen+1];
buffer[bufLen] = 0; // extra null at the end to protect strchr()
// to run over the buffer.
NABoolean sampleDone = FALSE;
Int32 totalSamples = 10;
Int32 totalLen = 0;
while (!sampleDone) {
tSize szRead = hdfsPread(fs, file, offset, buffer, bufLen);
char* pos = NULL;
if ( szRead > 0 ) {
//if (isSequenceFile && offset==0 && memcmp(buffer, "SEQ6", 4) == 0)
// isSequenceFile_ = TRUE;
char* start = buffer;
for (Int32 i=0; i<totalSamples; i++ ) {
if ( (pos=strchr(start, recordTerminator)) ) {
totalLen += pos - start + 1 + offset;
samples++;
start = pos+1;
if ( start > buffer + bufLen ) {
sampleDone = TRUE;
break;
}
} else
break;
}
if ( samples > 0 )
break;
else
offset += bufLen;
} else
break; // fail to read any bytes. Bail out.
}
NADELETEBASIC(buffer, heap_);
if ( samples > 0 ) {
sampledBytes_ += totalLen;
sampledRows_ += samples;
}
hdfsCloseFile(fs, file);
} else {
// can not do hdfs open on the file. Assume the file is empty.
}
}
if (blockSize_)
{
numBlocks_ = totalSize_ / blockSize_;
if (totalSize_ % blockSize_ > 0)
numBlocks_++; // partial block at the end
}
else
{
CMPASSERT(blockSize_);
// TBD:DIAGS
result = FALSE;
}
if ( totalSize_ > 0 )
{
blockHosts_ = new(heap_) HostId[replication_*numBlocks_];
// walk through blocks and record their locations
tOffset o = 0;
Int64 blockNum;
for (blockNum=0; blockNum < numBlocks_ && result; blockNum++)
{
char*** blockHostNames = hdfsGetHosts(fs,
fileInfo->mName,
o,
fileInfo->mBlockSize);
o += blockSize_;
if (blockHostNames == NULL)
{
CMPASSERT(blockHostNames);
// TBD:DIAGS
result = FALSE;
}
else
{
char **h = *blockHostNames;
HostId hostId;
for (Int32 r=0; r<replication_; r++)
{
if (h[r])
hostId = HHDFSMasterHostList::getHostNum(h[r]);
else
hostId = HHDFSMasterHostList::InvalidHostId;
blockHosts_[r*numBlocks_+blockNum] = hostId;
}
if (sortHosts)
sortHostArray(blockHosts_,
(Int32) numBlocks_,
replication_,
getFileName());
}
hdfsFreeHosts(blockHostNames);
}
}
return result;
}
int main(int argc, char **argv) {
const char *writePath = "/tmp/testfile.txt";
const char *fileContents = "Hello, World!";
const char *readPath = "/tmp/testfile.txt";
const char *srcPath = "/tmp/testfile.txt";
const char *dstPath = "/tmp/testfile2.txt";
const char *slashTmp = "/tmp";
const char *newDirectory = "/tmp/newdir";
const char *newOwner = "root";
const char *tuser = "******";
const char *appendPath = "/tmp/appends";
const char *userPath = "/tmp/usertestfile.txt";
char buffer[32], buffer2[256], rdbuffer[32];
tSize num_written_bytes, num_read_bytes;
hdfsFS fs, lfs;
hdfsFile writeFile, readFile, localFile, appendFile, userFile;
tOffset currentPos, seekPos;
int exists, totalResult, result, numEntries, i, j;
const char *resp;
hdfsFileInfo *fileInfo, *fileList, *finfo;
char *buffer3;
char permissions[10];
char ***hosts;
short newPerm = 0666;
tTime newMtime, newAtime;
fs = hdfsConnectNewInstance("default", 0);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs!\n");
exit(-1);
}
lfs = hdfsConnectNewInstance(NULL, 0);
if(!lfs) {
fprintf(stderr, "Oops! Failed to connect to 'local' hdfs!\n");
exit(-1);
}
{
//Write tests
writeFile = hdfsOpenFile(fs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", writePath);
num_written_bytes =
hdfsWrite(fs, writeFile, (void*)fileContents,
(tSize)(strlen(fileContents)+1));
if (num_written_bytes != strlen(fileContents) + 1) {
fprintf(stderr, "Failed to write correct number of bytes - expected %d, got %d\n",
(int)(strlen(fileContents) + 1), (int)num_written_bytes);
exit(-1);
}
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
currentPos = -1;
if ((currentPos = hdfsTell(fs, writeFile)) == -1) {
fprintf(stderr,
"Failed to get current file position correctly! Got %" PRId64 "!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %" PRId64 "\n", currentPos);
if (hdfsFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", writePath);
if (hdfsHFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'hflush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "HFlushed %s successfully!\n", writePath);
hdfsCloseFile(fs, writeFile);
}
{
//Read tests
exists = hdfsExists(fs, readPath);
if (exists) {
fprintf(stderr, "Failed to validate existence of %s\n", readPath);
exit(-1);
}
readFile = hdfsOpenFile(fs, readPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", readPath);
exit(-1);
}
if (!hdfsFileIsOpenForRead(readFile)) {
fprintf(stderr, "hdfsFileIsOpenForRead: we just opened a file "
"with O_RDONLY, and it did not show up as 'open for "
"read'\n");
exit(-1);
}
fprintf(stderr, "hdfsAvailable: %d\n", hdfsAvailable(fs, readFile));
seekPos = 1;
if(hdfsSeek(fs, readFile, seekPos)) {
fprintf(stderr, "Failed to seek %s for reading!\n", readPath);
exit(-1);
}
currentPos = -1;
if((currentPos = hdfsTell(fs, readFile)) != seekPos) {
fprintf(stderr,
"Failed to get current file position correctly! Got %" PRId64 "!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %" PRId64 "\n", currentPos);
if (!hdfsFileUsesDirectRead(readFile)) {
fprintf(stderr, "Direct read support incorrectly not detected "
"for HDFS filesystem\n");
exit(-1);
}
fprintf(stderr, "Direct read support detected for HDFS\n");
// Test the direct read path
if(hdfsSeek(fs, readFile, 0)) {
fprintf(stderr, "Failed to seek %s for reading!\n", readPath);
exit(-1);
}
memset(buffer, 0, sizeof(buffer));
num_read_bytes = hdfsRead(fs, readFile, (void*)buffer,
sizeof(buffer));
if (strncmp(fileContents, buffer, strlen(fileContents)) != 0) {
fprintf(stderr, "Failed to read (direct). Expected %s but got %s (%d bytes)\n",
fileContents, buffer, num_read_bytes);
exit(-1);
}
fprintf(stderr, "Read (direct) following %d bytes:\n%s\n",
num_read_bytes, buffer);
if (hdfsSeek(fs, readFile, 0L)) {
fprintf(stderr, "Failed to seek to file start!\n");
exit(-1);
}
// Disable the direct read path so that we really go through the slow
// read path
hdfsFileDisableDirectRead(readFile);
num_read_bytes = hdfsRead(fs, readFile, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
memset(buffer, 0, strlen(fileContents + 1));
num_read_bytes = hdfsPread(fs, readFile, 0, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
hdfsCloseFile(fs, readFile);
// Test correct behaviour for unsupported filesystems
localFile = hdfsOpenFile(lfs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!localFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
num_written_bytes = hdfsWrite(lfs, localFile, (void*)fileContents,
(tSize)(strlen(fileContents) + 1));
hdfsCloseFile(lfs, localFile);
localFile = hdfsOpenFile(lfs, writePath, O_RDONLY, 0, 0, 0);
if (hdfsFileUsesDirectRead(localFile)) {
fprintf(stderr, "Direct read support incorrectly detected for local "
"filesystem\n");
exit(-1);
}
hdfsCloseFile(lfs, localFile);
}
totalResult = 0;
result = 0;
{
//Generic file-system operations
fprintf(stderr, "hdfsCopy(remote-local): %s\n", ((result = hdfsCopy(fs, srcPath, lfs, srcPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(local-local): %s\n", ((result = hdfsMove(lfs, srcPath, lfs, dstPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(remote-local): %s\n", ((result = hdfsMove(fs, srcPath, lfs, srcPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsRename: %s\n", ((result = hdfsRename(fs, dstPath, srcPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCreateDirectory: %s\n", ((result = hdfsCreateDirectory(fs, newDirectory)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsSetReplication: %s\n", ((result = hdfsSetReplication(fs, srcPath, 2)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer2, sizeof(buffer2))) != 0 ? buffer2 : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsSetWorkingDirectory: %s\n", ((result = hdfsSetWorkingDirectory(fs, slashTmp)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer2, sizeof(buffer2))) != 0 ? buffer2 : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsGetDefaultBlockSize: %" PRId64 "\n", hdfsGetDefaultBlockSize(fs));
fprintf(stderr, "hdfsGetCapacity: %" PRId64 "\n", hdfsGetCapacity(fs));
fprintf(stderr, "hdfsGetUsed: %" PRId64 "\n", hdfsGetUsed(fs));
fileInfo = NULL;
if((fileInfo = hdfsGetPathInfo(fs, slashTmp)) != NULL) {
fprintf(stderr, "hdfsGetPathInfo - SUCCESS!\n");
fprintf(stderr, "Name: %s, ", fileInfo->mName);
fprintf(stderr, "Type: %c, ", (char)(fileInfo->mKind));
fprintf(stderr, "Replication: %d, ", fileInfo->mReplication);
fprintf(stderr, "BlockSize: %" PRId64 ", ", fileInfo->mBlockSize);
fprintf(stderr, "Size: %" PRId64 ", ", fileInfo->mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileInfo->mLastMod));
fprintf(stderr, "Owner: %s, ", fileInfo->mOwner);
fprintf(stderr, "Group: %s, ", fileInfo->mGroup);
permission_disp(fileInfo->mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileInfo->mPermissions, permissions);
hdfsFreeFileInfo(fileInfo, 1);
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetPathInfo for %s - FAILED!\n", slashTmp);
}
fileList = 0;
fileList = hdfsListDirectory(fs, newDirectory, &numEntries);
if (!(fileList == NULL && numEntries == 0 && !errno)) {
fprintf(stderr, "waah! hdfsListDirectory for empty %s - FAILED!\n", newDirectory);
totalResult++;
} else {
fprintf(stderr, "hdfsListDirectory for empty %s - SUCCESS!\n", newDirectory);
}
fileList = 0;
if((fileList = hdfsListDirectory(fs, slashTmp, &numEntries)) != NULL) {
for(i=0; i < numEntries; ++i) {
fprintf(stderr, "Name: %s, ", fileList[i].mName);
fprintf(stderr, "Type: %c, ", (char)fileList[i].mKind);
fprintf(stderr, "Replication: %d, ", fileList[i].mReplication);
fprintf(stderr, "BlockSize: %" PRId64 ", ", fileList[i].mBlockSize);
fprintf(stderr, "Size: %" PRId64 ", ", fileList[i].mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileList[i].mLastMod));
fprintf(stderr, "Owner: %s, ", fileList[i].mOwner);
fprintf(stderr, "Group: %s, ", fileList[i].mGroup);
permission_disp(fileList[i].mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileList[i].mPermissions, permissions);
}
hdfsFreeFileInfo(fileList, numEntries);
} else {
if (errno) {
totalResult++;
fprintf(stderr, "waah! hdfsListDirectory - FAILED!\n");
} else {
fprintf(stderr, "Empty directory!\n");
}
}
hosts = hdfsGetHosts(fs, srcPath, 0, 1);
if(hosts) {
fprintf(stderr, "hdfsGetHosts - SUCCESS! ... \n");
i=0;
while(hosts[i]) {
j = 0;
while(hosts[i][j]) {
fprintf(stderr,
"\thosts[%d][%d] - %s\n", i, j, hosts[i][j]);
++j;
}
++i;
}
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetHosts - FAILED!\n");
}
// setting tmp dir to 777 so later when connectAsUser nobody, we can write to it
// chown write
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, NULL, "users")) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, newOwner, NULL)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
// chmod write
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, writePath, newPerm)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
sleep(2);
newMtime = time(NULL);
newAtime = time(NULL);
// utime write
fprintf(stderr, "hdfsUtime: %s\n", ((result = hdfsUtime(fs, writePath, newMtime, newAtime)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
// chown/chmod/utime read
finfo = hdfsGetPathInfo(fs, writePath);
fprintf(stderr, "hdfsChown read: %s\n", ((result = (strcmp(finfo->mOwner, newOwner))) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChmod read: %s\n", ((result = (finfo->mPermissions != newPerm)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
// will later use /tmp/ as a different user so enable it
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, "/tmp/", 0777)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr,"newMTime=%ld\n",newMtime);
fprintf(stderr,"curMTime=%ld\n",finfo->mLastMod);
fprintf(stderr, "hdfsUtime read (mtime): %s\n", ((result = (finfo->mLastMod != newMtime)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
// No easy way to turn on access times from hdfs_test right now
// fprintf(stderr, "hdfsUtime read (atime): %s\n", ((result = (finfo->mLastAccess != newAtime)) != 0 ? "Failed!" : "Success!"));
// totalResult += result;
hdfsFreeFileInfo(finfo, 1);
// Clean up
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, newDirectory, 1)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, srcPath, 1)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, srcPath, 1)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, dstPath, 1)) != 0 ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsExists: %s\n", ((result = hdfsExists(fs, newDirectory)) != 0 ? "Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
}
{
// TEST APPENDS
// CREATE
appendFile = hdfsOpenFile(fs, appendPath, O_WRONLY, 0, 0, 0);
if(!appendFile) {
fprintf(stderr, "Failed to open %s for writing!\n", appendPath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", appendPath);
buffer3 = "Hello,";
num_written_bytes = hdfsWrite(fs, appendFile, (void*)buffer3,
(tSize)strlen(buffer3));
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
if (hdfsFlush(fs, appendFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", appendPath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", appendPath);
hdfsCloseFile(fs, appendFile);
// RE-OPEN
appendFile = hdfsOpenFile(fs, appendPath, O_WRONLY|O_APPEND, 0, 0, 0);
if(!appendFile) {
fprintf(stderr, "Failed to open %s for writing!\n", appendPath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", appendPath);
buffer3 = " World";
num_written_bytes = hdfsWrite(fs, appendFile, (void*)buffer3,
(tSize)(strlen(buffer3) + 1));
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
if (hdfsFlush(fs, appendFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", appendPath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", appendPath);
hdfsCloseFile(fs, appendFile);
// CHECK size
finfo = hdfsGetPathInfo(fs, appendPath);
fprintf(stderr, "fileinfo->mSize: == total %s\n", ((result = (finfo->mSize == (tOffset)(strlen("Hello, World") + 1))) == 1 ? "Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
// READ and check data
readFile = hdfsOpenFile(fs, appendPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", appendPath);
exit(-1);
}
num_read_bytes = hdfsRead(fs, readFile, (void*)rdbuffer, sizeof(rdbuffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, rdbuffer);
fprintf(stderr, "read == Hello, World %s\n", ((result = (strcmp(rdbuffer, "Hello, World"))) == 0 ? "Success!" : "Failed!"));
hdfsCloseFile(fs, readFile);
// DONE test appends
}
totalResult += (hdfsDisconnect(fs) != 0);
{
//
// Now test as connecting as a specific user
// This is only meant to test that we connected as that user, not to test
// the actual fs user capabilities. Thus just create a file and read
// the owner is correct.
fs = hdfsConnectAsUserNewInstance("default", 0, tuser);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs as user %s!\n",tuser);
exit(-1);
}
userFile = hdfsOpenFile(fs, userPath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!userFile) {
fprintf(stderr, "Failed to open %s for writing!\n", userPath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", userPath);
num_written_bytes = hdfsWrite(fs, userFile, (void*)fileContents,
(tSize)(strlen(fileContents)+1));
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
if (hdfsFlush(fs, userFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", userPath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", userPath);
hdfsCloseFile(fs, userFile);
finfo = hdfsGetPathInfo(fs, userPath);
fprintf(stderr, "hdfs new file user is correct: %s\n", ((result = (strcmp(finfo->mOwner, tuser))) != 0 ? "Failed!" : "Success!"));
totalResult += result;
}
totalResult += (hdfsDisconnect(fs) != 0);
if (totalResult != 0) {
return -1;
} else {
return 0;
}
}
int main(int argc, char **argv)
{
char buffer[32];
tSize num_written_bytes;
const char* slashTmp = "/tmp";
int nnPort;
char *rwTemplate, *rwTemplate2, *newDirTemplate,
*appendTemplate, *userTemplate, *rwPath = NULL;
const char* fileContents = "Hello, World!";
const char* nnHost = NULL;
if (argc != 2) {
fprintf(stderr, "usage: test_libwebhdfs_ops <username>\n");
exit(1);
}
struct NativeMiniDfsConf conf = {
.doFormat = 1, .webhdfsEnabled = 1, .namenodeHttpPort = 50070,
};
cluster = nmdCreate(&conf);
if (!cluster) {
fprintf(stderr, "Failed to create the NativeMiniDfsCluster.\n");
exit(1);
}
if (nmdWaitClusterUp(cluster)) {
fprintf(stderr, "Error when waiting for cluster to be ready.\n");
exit(1);
}
if (nmdGetNameNodeHttpAddress(cluster, &nnPort, &nnHost)) {
fprintf(stderr, "Error when retrieving namenode host address.\n");
exit(1);
}
hdfsFS fs = hdfsConnectAsUserNewInstance(nnHost, nnPort, argv[1]);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs!\n");
exit(-1);
}
{
// Write tests
rwTemplate = strdup("/tmp/helloWorldXXXXXX");
if (!rwTemplate) {
fprintf(stderr, "Failed to create rwTemplate!\n");
exit(1);
}
rwPath = mktemp(rwTemplate);
// hdfsOpenFile
hdfsFile writeFile = hdfsOpenFile(fs, rwPath,
O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", rwPath);
exit(1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", rwPath);
// hdfsWrite
num_written_bytes = hdfsWrite(fs, writeFile, (void*)fileContents,
(int) strlen(fileContents) + 1);
if (num_written_bytes != strlen(fileContents) + 1) {
fprintf(stderr, "Failed to write correct number of bytes - "
"expected %d, got %d\n",
(int)(strlen(fileContents) + 1), (int) num_written_bytes);
exit(1);
}
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
// hdfsTell
tOffset currentPos = -1;
if ((currentPos = hdfsTell(fs, writeFile)) == -1) {
fprintf(stderr,
"Failed to get current file position correctly. Got %"
PRId64 "!\n", currentPos);
exit(1);
}
fprintf(stderr, "Current position: %" PRId64 "\n", currentPos);
hdfsCloseFile(fs, writeFile);
// Done test write
}
sleep(1);
{
//Read tests
int available = 0, exists = 0;
// hdfsExists
exists = hdfsExists(fs, rwPath);
if (exists) {
fprintf(stderr, "Failed to validate existence of %s\n", rwPath);
exists = hdfsExists(fs, rwPath);
if (exists) {
fprintf(stderr,
"Still failed to validate existence of %s\n", rwPath);
exit(1);
}
}
hdfsFile readFile = hdfsOpenFile(fs, rwPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", rwPath);
exit(1);
}
if (!hdfsFileIsOpenForRead(readFile)) {
fprintf(stderr, "hdfsFileIsOpenForRead: we just opened a file "
"with O_RDONLY, and it did not show up as 'open for "
"read'\n");
exit(1);
}
available = hdfsAvailable(fs, readFile);
fprintf(stderr, "hdfsAvailable: %d\n", available);
// hdfsSeek, hdfsTell
tOffset seekPos = 1;
if(hdfsSeek(fs, readFile, seekPos)) {
fprintf(stderr, "Failed to seek %s for reading!\n", rwPath);
exit(1);
}
tOffset currentPos = -1;
if((currentPos = hdfsTell(fs, readFile)) != seekPos) {
fprintf(stderr,
"Failed to get current file position correctly! Got %"
PRId64 "!\n", currentPos);
exit(1);
}
fprintf(stderr, "Current position: %" PRId64 "\n", currentPos);
if(hdfsSeek(fs, readFile, 0)) {
fprintf(stderr, "Failed to seek %s for reading!\n", rwPath);
exit(1);
}
// hdfsRead
memset(buffer, 0, sizeof(buffer));
tSize num_read_bytes = hdfsRead(fs, readFile, buffer, sizeof(buffer));
if (strncmp(fileContents, buffer, strlen(fileContents)) != 0) {
fprintf(stderr, "Failed to read (direct). "
"Expected %s but got %s (%d bytes)\n",
fileContents, buffer, num_read_bytes);
exit(1);
}
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
if (hdfsSeek(fs, readFile, 0L)) {
fprintf(stderr, "Failed to seek to file start!\n");
exit(1);
}
// hdfsPread
memset(buffer, 0, strlen(fileContents + 1));
num_read_bytes = hdfsPread(fs, readFile, 0, buffer, sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
hdfsCloseFile(fs, readFile);
// Done test read
}
int totalResult = 0;
int result = 0;
{
//Generic file-system operations
char *srcPath = rwPath;
char buffer[256];
const char *resp;
rwTemplate2 = strdup("/tmp/helloWorld2XXXXXX");
if (!rwTemplate2) {
fprintf(stderr, "Failed to create rwTemplate2!\n");
exit(1);
}
char *dstPath = mktemp(rwTemplate2);
newDirTemplate = strdup("/tmp/newdirXXXXXX");
if (!newDirTemplate) {
fprintf(stderr, "Failed to create newDirTemplate!\n");
exit(1);
}
char *newDirectory = mktemp(newDirTemplate);
// hdfsRename
fprintf(stderr, "hdfsRename: %s\n",
((result = hdfsRename(fs, rwPath, dstPath)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsRename back: %s\n",
((result = hdfsRename(fs, dstPath, srcPath)) ?
"Failed!" : "Success!"));
totalResult += result;
// hdfsCreateDirectory
fprintf(stderr, "hdfsCreateDirectory: %s\n",
((result = hdfsCreateDirectory(fs, newDirectory)) ?
"Failed!" : "Success!"));
totalResult += result;
// hdfsSetReplication
fprintf(stderr, "hdfsSetReplication: %s\n",
((result = hdfsSetReplication(fs, srcPath, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
// hdfsGetWorkingDirectory, hdfsSetWorkingDirectory
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n",
((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ?
buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
const char* path[] = {"/foo", "/foo/bar", "foobar", "//foo/bar//foobar",
"foo//bar", "foo/bar///", "/", "////"};
int i;
for (i = 0; i < 8; i++) {
fprintf(stderr, "hdfsSetWorkingDirectory: %s, %s\n",
((result = hdfsSetWorkingDirectory(fs, path[i])) ?
"Failed!" : "Success!"),
hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer)));
totalResult += result;
}
fprintf(stderr, "hdfsSetWorkingDirectory: %s\n",
((result = hdfsSetWorkingDirectory(fs, slashTmp)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n",
((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ?
buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
// hdfsGetPathInfo
hdfsFileInfo *fileInfo = NULL;
if((fileInfo = hdfsGetPathInfo(fs, slashTmp)) != NULL) {
fprintf(stderr, "hdfsGetPathInfo - SUCCESS!\n");
fprintf(stderr, "Name: %s, ", fileInfo->mName);
fprintf(stderr, "Type: %c, ", (char)(fileInfo->mKind));
fprintf(stderr, "Replication: %d, ", fileInfo->mReplication);
fprintf(stderr, "BlockSize: %"PRId64", ", fileInfo->mBlockSize);
fprintf(stderr, "Size: %"PRId64", ", fileInfo->mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileInfo->mLastMod));
fprintf(stderr, "Owner: %s, ", fileInfo->mOwner);
fprintf(stderr, "Group: %s, ", fileInfo->mGroup);
char permissions[10];
permission_disp(fileInfo->mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n",
fileInfo->mPermissions, permissions);
hdfsFreeFileInfo(fileInfo, 1);
} else {
totalResult++;
fprintf(stderr, "hdfsGetPathInfo for %s - FAILED!\n", slashTmp);
}
// hdfsListDirectory
hdfsFileInfo *fileList = 0;
int numEntries = 0;
if((fileList = hdfsListDirectory(fs, slashTmp, &numEntries)) != NULL) {
int i = 0;
for(i=0; i < numEntries; ++i) {
fprintf(stderr, "Name: %s, ", fileList[i].mName);
fprintf(stderr, "Type: %c, ", (char)fileList[i].mKind);
fprintf(stderr, "Replication: %d, ", fileList[i].mReplication);
fprintf(stderr, "BlockSize: %"PRId64", ", fileList[i].mBlockSize);
fprintf(stderr, "Size: %"PRId64", ", fileList[i].mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileList[i].mLastMod));
fprintf(stderr, "Owner: %s, ", fileList[i].mOwner);
fprintf(stderr, "Group: %s, ", fileList[i].mGroup);
char permissions[10];
permission_disp(fileList[i].mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n",
fileList[i].mPermissions, permissions);
}
hdfsFreeFileInfo(fileList, numEntries);
} else {
if (errno) {
totalResult++;
fprintf(stderr, "waah! hdfsListDirectory - FAILED!\n");
} else {
fprintf(stderr, "Empty directory!\n");
}
}
char *newOwner = "root";
// Setting tmp dir to 777 so later when connectAsUser nobody,
// we can write to it
short newPerm = 0666;
// hdfsChown
fprintf(stderr, "hdfsChown: %s\n",
((result = hdfsChown(fs, rwPath, NULL, "users")) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChown: %s\n",
((result = hdfsChown(fs, rwPath, newOwner, NULL)) ?
"Failed!" : "Success!"));
totalResult += result;
// hdfsChmod
fprintf(stderr, "hdfsChmod: %s\n",
((result = hdfsChmod(fs, rwPath, newPerm)) ?
"Failed!" : "Success!"));
totalResult += result;
sleep(2);
tTime newMtime = time(NULL);
tTime newAtime = time(NULL);
// utime write
fprintf(stderr, "hdfsUtime: %s\n",
((result = hdfsUtime(fs, rwPath, newMtime, newAtime)) ?
"Failed!" : "Success!"));
totalResult += result;
// chown/chmod/utime read
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, rwPath);
fprintf(stderr, "hdfsChown read: %s\n",
((result = (strcmp(finfo->mOwner, newOwner) != 0)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChmod read: %s\n",
((result = (finfo->mPermissions != newPerm)) ?
"Failed!" : "Success!"));
totalResult += result;
// will later use /tmp/ as a different user so enable it
fprintf(stderr, "hdfsChmod: %s\n",
((result = hdfsChmod(fs, slashTmp, 0777)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr,"newMTime=%ld\n",newMtime);
fprintf(stderr,"curMTime=%ld\n",finfo->mLastMod);
fprintf(stderr, "hdfsUtime read (mtime): %s\n",
((result = (finfo->mLastMod != newMtime / 1000)) ?
"Failed!" : "Success!"));
totalResult += result;
// Clean up
hdfsFreeFileInfo(finfo, 1);
fprintf(stderr, "hdfsDelete: %s\n",
((result = hdfsDelete(fs, newDirectory, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n",
((result = hdfsDelete(fs, srcPath, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsExists: %s\n",
((result = hdfsExists(fs, newDirectory)) ?
"Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
// Done test generic operations
}
{
// Test Appends
appendTemplate = strdup("/tmp/appendsXXXXXX");
if (!appendTemplate) {
fprintf(stderr, "Failed to create appendTemplate!\n");
exit(1);
}
char *appendPath = mktemp(appendTemplate);
const char* helloBuffer = "Hello,";
hdfsFile writeFile = NULL;
// Create
writeFile = hdfsOpenFile(fs, appendPath, O_WRONLY, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", appendPath);
exit(1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", appendPath);
num_written_bytes = hdfsWrite(fs, writeFile, helloBuffer,
(int) strlen(helloBuffer));
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
hdfsCloseFile(fs, writeFile);
fprintf(stderr, "hdfsSetReplication: %s\n",
((result = hdfsSetReplication(fs, appendPath, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
// Re-Open for Append
writeFile = hdfsOpenFile(fs, appendPath, O_WRONLY | O_APPEND, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", appendPath);
exit(1);
}
fprintf(stderr, "Opened %s for appending successfully...\n",
appendPath);
helloBuffer = " World";
num_written_bytes = hdfsWrite(fs, writeFile, helloBuffer,
(int)strlen(helloBuffer) + 1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
hdfsCloseFile(fs, writeFile);
// Check size
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, appendPath);
fprintf(stderr, "fileinfo->mSize: == total %s\n",
((result = (finfo->mSize == strlen("Hello, World") + 1)) ?
"Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
// Read and check data
hdfsFile readFile = hdfsOpenFile(fs, appendPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", appendPath);
exit(1);
}
tSize num_read_bytes = hdfsRead(fs, readFile, buffer, sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
fprintf(stderr, "read == Hello, World %s\n",
(result = (strcmp(buffer, "Hello, World") == 0)) ?
"Success!" : "Failed!");
hdfsCloseFile(fs, readFile);
// Cleanup
fprintf(stderr, "hdfsDelete: %s\n",
((result = hdfsDelete(fs, appendPath, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
// Done test appends
}
totalResult += (hdfsDisconnect(fs) != 0);
{
//
// Now test as connecting as a specific user
// This only meant to test that we connected as that user, not to test
// the actual fs user capabilities. Thus just create a file and read
// the owner is correct.
const char *tuser = "******";
userTemplate = strdup("/tmp/usertestXXXXXX");
if (!userTemplate) {
fprintf(stderr, "Failed to create userTemplate!\n");
exit(1);
}
char* userWritePath = mktemp(userTemplate);
hdfsFile writeFile = NULL;
fs = hdfsConnectAsUserNewInstance("default", 50070, tuser);
if(!fs) {
fprintf(stderr,
"Oops! Failed to connect to hdfs as user %s!\n",tuser);
exit(1);
}
writeFile = hdfsOpenFile(fs, userWritePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", userWritePath);
exit(1);
}
fprintf(stderr, "Opened %s for writing successfully...\n",
userWritePath);
num_written_bytes = hdfsWrite(fs, writeFile, fileContents,
(int)strlen(fileContents) + 1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
hdfsCloseFile(fs, writeFile);
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, userWritePath);
if (finfo) {
fprintf(stderr, "hdfs new file user is correct: %s\n",
((result = (strcmp(finfo->mOwner, tuser) != 0)) ?
"Failed!" : "Success!"));
} else {
fprintf(stderr,
"hdfsFileInfo returned by hdfsGetPathInfo is NULL\n");
result = -1;
}
totalResult += result;
// Cleanup
fprintf(stderr, "hdfsDelete: %s\n",
((result = hdfsDelete(fs, userWritePath, 1)) ?
"Failed!" : "Success!"));
totalResult += result;
// Done test specific user
}
totalResult += (hdfsDisconnect(fs) != 0);
// Shutdown the native minidfscluster
nmdShutdown(cluster);
nmdFree(cluster);
fprintf(stderr, "totalResult == %d\n", totalResult);
if (totalResult != 0) {
return -1;
} else {
return 0;
}
}
qioerr hdfs_preadv (void* file, const struct iovec *vector, int count, off_t offset, ssize_t* num_read_out, void* fs)
{
ssize_t got;
ssize_t got_total;
qioerr err_out = 0;
int i;
STARTING_SLOW_SYSCALL;
#ifdef HDFS3
const hdfs_file orig_hfl = *to_hdfs_file(file);
const hdfs_fs orig_hfs = *to_hdfs_fs(fs);
hdfsFS hfs = hdfsConnect(orig_hfs.fs_name, orig_hfs.fs_port);
hdfsFile hfl = hdfsOpenFile(hfs, orig_hfl.pathnm, O_RDONLY, 0, 0, 0);
//assert connection
CREATE_ERROR((hfs == NULL), err_out, ECONNREFUSED, "Unable to read HDFS file", error);
if(hfl == NULL) {
err_out = qio_mkerror_errno();
goto error;
}
#endif
err_out = 0;
got_total = 0;
for(i = 0; i < count; i++) {
#ifdef HDFS3
hdfsSeek(hfs, hfl, offset+got_total);
got = hdfsRead(hfs, hfl, (void*)vector[i].iov_base, vector[i].iov_len);
#else
got = hdfsPread(to_hdfs_fs(fs)->hfs, to_hdfs_file(file)->file, offset + got_total, (void*)vector[i].iov_base, vector[i].iov_len);
#endif
if( got != -1 ) {
got_total += got;
} else {
err_out = qio_mkerror_errno();
break;
}
if(got != (ssize_t)vector[i].iov_len ) {
break;
}
}
if( err_out == 0 && got_total == 0 && sys_iov_total_bytes(vector, count) != 0 )
err_out = qio_int_to_err(EEOF);
*num_read_out = got_total;
#ifdef HDFS3
got = hdfsCloseFile(hfs, hfl);
if(got == -1) { err_out = qio_mkerror_errno(); }
got = hdfsDisconnect(hfs);
if(got == -1) { err_out = qio_mkerror_errno(); }
#endif
DONE_SLOW_SYSCALL;
#ifdef HDFS3
error:
#endif
return err_out;
}
