vitalif
/
mdtest
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
mdtest/src/aiori-HDFS.c

698 lines
20 KiB
C
Executable File
Raw Permalink Blame History

/* -*- mode: c; indent-tabs-mode: t; -*-
* vim:noexpandtab:
*
* Editing with tabs allows different users to pick their own indentation
* appearance without changing the file.
*/
/*
* Copyright (c) 2009, Los Alamos National Security, LLC All rights reserved.
* Copyright 2009. Los Alamos National Security, LLC. This software was produced
* under U.S. Government contract DE-AC52-06NA25396 for Los Alamos National
* Laboratory (LANL), which is operated by Los Alamos National Security, LLC for
* the U.S. Department of Energy. The U.S. Government has rights to use,
* reproduce, and distribute this software. NEITHER THE GOVERNMENT NOR LOS
* ALAMOS NATIONAL SECURITY, LLC MAKES ANY WARRANTY, EXPRESS OR IMPLIED, OR
* ASSUMES ANY LIABILITY FOR THE USE OF THIS SOFTWARE. If software is
* modified to produce derivative works, such modified software should be
* clearly marked, so as not to confuse it with the version available from
* LANL.
*
* Additionally, redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following conditions are
* met:
*
* • Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* • Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* • Neither the name of Los Alamos National Security, LLC, Los Alamos National
* Laboratory, LANL, the U.S. Government, nor the names of its contributors may be
* used to endorse or promote products derived from this software without specific
* prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY LOS ALAMOS NATIONAL SECURITY, LLC AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
* THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL LOS ALAMOS NATIONAL SECURITY, LLC OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT
* OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
* OF SUCH DAMAGE.
*/
/******************************************************************************\
*
* Implement of abstract I/O interface for HDFS.
*
* HDFS has the added concept of a "File System Handle" which has to be
* connected before files are opened. We store this in the IOR_param_t
* object that is always passed to our functions. The thing that callers
* think of as the "fd" is an hdfsFile, (a pointer).
*
\******************************************************************************/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
#include <stdio.h>
#include <stdlib.h>
#ifdef __linux__
# include <sys/ioctl.h> /* necessary for: */
# define __USE_GNU /* O_DIRECT and */
# include <fcntl.h> /* IO operations */
# undef __USE_GNU
#endif /* __linux__ */
#include <errno.h>
#include <fcntl.h> /* IO operations */
#include <sys/stat.h>
#include <assert.h>
/*
#ifdef HAVE_LUSTRE_USER
#include <lustre/lustre_user.h>
#endif
*/
#include "ior.h"
#include "aiori.h"
#include "utilities.h"
#ifndef open64 /* necessary for TRU64 -- */
# define open64 open /* unlikely, but may pose */
#endif /* not open64 */ /* conflicting prototypes */
#ifndef lseek64 /* necessary for TRU64 -- */
# define lseek64 lseek /* unlikely, but may pose */
#endif /* not lseek64 */ /* conflicting prototypes */
#ifndef O_BINARY /* Required on Windows */
# define O_BINARY 0
#endif
#include "hdfs.h"
/**************************** P R O T O T Y P E S *****************************/
static aiori_fd_t *HDFS_Create(char *testFileName, int flags, aiori_mod_opt_t * param);
static aiori_fd_t *HDFS_Open(char *testFileName, int flags, aiori_mod_opt_t * param);
static IOR_offset_t HDFS_Xfer(int access, aiori_fd_t *file, IOR_size_t * buffer,
IOR_offset_t length, IOR_offset_t offset, aiori_mod_opt_t * param);
static void HDFS_Close(aiori_fd_t *, aiori_mod_opt_t *);
static void HDFS_Delete(char *testFileName, aiori_mod_opt_t * param);
static void HDFS_Fsync(aiori_fd_t *, aiori_mod_opt_t *);
static IOR_offset_t HDFS_GetFileSize(aiori_mod_opt_t *,char *);
static void hdfs_xfer_hints(aiori_xfer_hint_t * params);
static option_help * HDFS_options(aiori_mod_opt_t ** init_backend_options, aiori_mod_opt_t * init_values);
static int HDFS_mkdir (const char *path, mode_t mode, aiori_mod_opt_t * options);
static int HDFS_rmdir (const char *path, aiori_mod_opt_t * options);
static int HDFS_access (const char *path, int mode, aiori_mod_opt_t * options);
static int HDFS_stat (const char *path, struct stat *buf, aiori_mod_opt_t * options);
static int HDFS_statfs (const char * path, ior_aiori_statfs_t * stat, aiori_mod_opt_t * options);
static aiori_xfer_hint_t * hints = NULL;
/************************** D E C L A R A T I O N S ***************************/
ior_aiori_t hdfs_aiori = {
.name = "HDFS",
.name_legacy = NULL,
.create = HDFS_Create,
.open = HDFS_Open,
.xfer = HDFS_Xfer,
.close = HDFS_Close,
.delete = HDFS_Delete,
.get_options = HDFS_options,
.get_version = aiori_get_version,
.xfer_hints = hdfs_xfer_hints,
.fsync = HDFS_Fsync,
.get_file_size = HDFS_GetFileSize,
.statfs = HDFS_statfs,
.mkdir = HDFS_mkdir,
.rmdir = HDFS_rmdir,
.access = HDFS_access,
.stat = HDFS_stat,
.enable_mdtest = true
};
/***************************** F U N C T I O N S ******************************/
void hdfs_xfer_hints(aiori_xfer_hint_t * params){
hints = params;
}
/************************** O P T I O N S *****************************/
typedef struct {
char * user;
char * name_node;
int replicas; /* n block replicas. (0 gets default) */
int direct_io;
IOR_offset_t block_size; /* internal blk-size. (0 gets default) */
// runtime options
hdfsFS fs; /* file-system handle */
tPort name_node_port; /* (uint16_t) */
} hdfs_options_t;
static void hdfs_connect( hdfs_options_t* o );
option_help * HDFS_options(aiori_mod_opt_t ** init_backend_options, aiori_mod_opt_t * init_values){
hdfs_options_t * o = malloc(sizeof(hdfs_options_t));
if (init_values != NULL){
memcpy(o, init_values, sizeof(hdfs_options_t));
}else{
memset(o, 0, sizeof(hdfs_options_t));
char *hdfs_user;
hdfs_user = getenv("USER");
if (!hdfs_user){
hdfs_user = "";
}
o->user = strdup(hdfs_user);
o->name_node = "default";
}
*init_backend_options = (aiori_mod_opt_t*) o;
option_help h [] = {
{0, "hdfs.odirect", "Direct I/O Mode", OPTION_FLAG, 'd', & o->direct_io},
{0, "hdfs.user", "Username", OPTION_OPTIONAL_ARGUMENT, 's', & o->user},
{0, "hdfs.name_node", "Namenode", OPTION_OPTIONAL_ARGUMENT, 's', & o->name_node},
{0, "hdfs.replicas", "Number of replicas", OPTION_OPTIONAL_ARGUMENT, 'd', & o->replicas},
{0, "hdfs.block_size", "Blocksize", OPTION_OPTIONAL_ARGUMENT, 'l', & o->block_size},
LAST_OPTION
};
option_help * help = malloc(sizeof(h));
memcpy(help, h, sizeof(h));
return help;
}
int HDFS_mkdir (const char *path, mode_t mode, aiori_mod_opt_t * options){
hdfs_options_t * o = (hdfs_options_t*) options;
hdfs_connect(o);
return hdfsCreateDirectory(o->fs, path);
}
int HDFS_rmdir (const char *path, aiori_mod_opt_t * options){
hdfs_options_t * o = (hdfs_options_t*) options;
hdfs_connect(o);
return hdfsDelete(o->fs, path, 1);
}
int HDFS_access (const char *path, int mode, aiori_mod_opt_t * options){
hdfs_options_t * o = (hdfs_options_t*) options;
hdfs_connect(o);
return hdfsExists(o->fs, path);
}
int HDFS_stat (const char *path, struct stat *buf, aiori_mod_opt_t * options){
hdfsFileInfo * stat;
hdfs_options_t * o = (hdfs_options_t*) options;
hdfs_connect(o);
stat = hdfsGetPathInfo(o->fs, path);
if(stat == NULL){
return 1;
}
memset(buf, 0, sizeof(struct stat));
buf->st_atime = stat->mLastAccess;
buf->st_size = stat->mSize;
buf->st_mtime = stat->mLastMod;
buf->st_mode = stat->mPermissions;
hdfsFreeFileInfo(stat, 1);
return 0;
}
int HDFS_statfs (const char * path, ior_aiori_statfs_t * stat, aiori_mod_opt_t * options){
hdfs_options_t * o = (hdfs_options_t*) options;
hdfs_connect(o);
stat->f_bsize = hdfsGetDefaultBlockSize(o->fs);
stat->f_blocks = hdfsGetCapacity(o->fs) / hdfsGetDefaultBlockSize(o->fs);
stat->f_bfree = stat->f_blocks - hdfsGetUsed(o->fs) / hdfsGetDefaultBlockSize(o->fs);
stat->f_bavail = 1;
stat->f_files = 1;
stat->f_ffree = 1;
return 0;
}
/* This is identical to the one in aiori-POSIX.c Doesn't seem like
* it would be appropriate in utilities.c.
*/
void hdfs_set_o_direct_flag(int *fd)
{
/* note that TRU64 needs O_DIRECTIO, SunOS uses directio(),
and everyone else needs O_DIRECT */
#ifndef O_DIRECT
# ifndef O_DIRECTIO
WARN("cannot use O_DIRECT");
# define O_DIRECT 000000
# else /* O_DIRECTIO */
# define O_DIRECT O_DIRECTIO
# endif /* not O_DIRECTIO */
#endif /* not O_DIRECT */
*fd |= O_DIRECT;
}
/*
* "Connect" to an HDFS file-system. HDFS requires this be done before and
* files are opened. It is easy for ior_aiori.open/create to assure that
* we connect, if we haven't already done so. However, there's not a
* simple way to assure that we disconnect after the last file-close. For
* now, we'll make a special call at the end of ior.c
*
* NOTE: It's okay to call this thing whenever you need to be sure the HDFS
* filesystem is connected.
*/
void hdfs_connect( hdfs_options_t* o ) {
if (verbose >= VERBOSE_4) {
printf("-> hdfs_connect [nn:\"%s\", port:%d, user:%s]\n",
o->name_node,
o->name_node_port,
o->user );
}
if ( o->fs ) {
if (verbose >= VERBOSE_4) {
printf("<- hdfs_connect [nothing to do]\n"); /* DEBUGGING */
}
return;
}
/* initialize a builder, holding parameters for hdfsBuilderConnect() */
struct hdfsBuilder* builder = hdfsNewBuilder();
if ( ! builder ){
ERR("couldn't create an hdfsBuilder");
}
hdfsBuilderSetForceNewInstance ( builder ); /* don't use cached instance */
hdfsBuilderSetNameNode ( builder, o->name_node );
hdfsBuilderSetNameNodePort( builder, o->name_node_port );
hdfsBuilderSetUserName ( builder, o->user );
/* NOTE: hdfsBuilderConnect() frees the builder */
o->fs = hdfsBuilderConnect( builder );
if ( ! o->fs )
ERR("hdsfsBuilderConnect failed");
if (verbose >= VERBOSE_4) {
printf("<- hdfs_connect [success]\n");
}
}
static void hdfs_disconnect( hdfs_options_t* o ) {
if (verbose >= VERBOSE_4) {
printf("-> hdfs_disconnect\n");
}
if ( o->fs ) {
hdfsDisconnect( o->fs );
o->fs = NULL;
}
if (verbose >= VERBOSE_4) {
printf("<- hdfs_disconnect\n");
}
}
/*
* Create or open the file. Pass TRUE if creating and FALSE if opening an existing file.
* Return an hdfsFile.
*/
static void *HDFS_Create_Or_Open( char *testFileName, int flags, aiori_mod_opt_t *param, unsigned char createFile ) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Create_Or_Open\n");
}
hdfs_options_t * o = (hdfs_options_t*) param;
hdfsFile hdfs_file = NULL;
int fd_oflags = 0, hdfs_return;
/* initialize file-system handle, if needed */
hdfs_connect( o );
/*
* Check for unsupported flags.
*
* If they want RDWR, we don't know if they're going to try to do both, so we
* can't default to either O_RDONLY or O_WRONLY. Thus, we error and exit.
*
* The other two, we just note that they are not supported and don't do them.
*/
if ( flags & IOR_RDWR ) {
ERR( "Opening or creating a file in RDWR is not implemented in HDFS" );
}
if ( flags & IOR_EXCL ) {
fprintf( stdout, "Opening or creating a file in Exclusive mode is not implemented in HDFS\n" );
}
if ( flags & IOR_APPEND ) {
fprintf( stdout, "Opening or creating a file for appending is not implemented in HDFS\n" );
}
/*
* Setup the flags to be used.
*/
if ( createFile == TRUE ) {
fd_oflags = O_CREAT;
}
if ( flags & IOR_WRONLY ) {
if ( ! hints->filePerProc ) {
// in N-1 mode, only rank 0 truncates the file
if ( rank != 0 ) {
fd_oflags |= O_WRONLY;
} else {
fd_oflags |= O_TRUNC;
fd_oflags |= O_WRONLY;
}
} else {
// in N-N mode, everyone does truncate
fd_oflags |= O_TRUNC;
fd_oflags |= O_WRONLY;
}
} else {
fd_oflags |= O_RDONLY;
}
/*
* Now see if O_DIRECT is needed.
*/
if ( o->direct_io == TRUE ) {
hdfs_set_o_direct_flag( &fd_oflags );
}
/*
* For N-1 write, All other ranks wait for Rank 0 to open the file.
* this is bec 0 does truncate and rest don't
* it would be dangerous for all to do truncate as they might
* truncate each other's writes
*/
if (( flags & IOR_WRONLY ) && ( ! hints->filePerProc ) && ( rank != 0 )) {
MPI_CHECK(MPI_Barrier(testComm), "barrier error");
}
/*
* Now rank zero can open and truncate, if necessary.
*/
if (verbose >= VERBOSE_4) {
printf("\thdfsOpenFile(%p, %s, 0%o, %lld, %d, %lld)\n",
o->fs,
testFileName,
fd_oflags, /* shown in octal to compare w/ <bits/fcntl.h> */
hints->transferSize,
o->replicas,
o->block_size);
}
hdfs_file = hdfsOpenFile( o->fs, testFileName, fd_oflags, hints->transferSize, o->replicas, o->block_size);
if ( ! hdfs_file ) {
ERR( "Failed to open the file" );
}
/*
* For N-1 write, Rank 0 waits for the other ranks to open the file after it has.
*/
if (( flags & IOR_WRONLY ) &&
( !hints->filePerProc ) &&
( rank == 0 )) {
MPI_CHECK(MPI_Barrier(testComm), "barrier error");
}
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Create_Or_Open\n");
}
return ((void *) hdfs_file );
}
/*
* Create and open a file through the HDFS interface.
*/
static aiori_fd_t *HDFS_Create(char *testFileName, int flags, aiori_mod_opt_t * param) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Create\n");
}
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Create\n");
}
return HDFS_Create_Or_Open( testFileName, flags, param, TRUE );
}
/*
* Open a file through the HDFS interface.
*/
static aiori_fd_t *HDFS_Open(char *testFileName, int flags, aiori_mod_opt_t * param) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Open\n");
}
if ( flags & IOR_CREAT ) {
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Open( ... TRUE)\n");
}
return HDFS_Create_Or_Open( testFileName, flags, param, TRUE );
}
else {
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Open( ... FALSE)\n");
}
return HDFS_Create_Or_Open( testFileName, flags, param, FALSE );
}
}
/*
* Write or read to file using the HDFS interface.
*/
static IOR_offset_t HDFS_Xfer(int access, aiori_fd_t *file, IOR_size_t * buffer,
IOR_offset_t length, IOR_offset_t offset, aiori_mod_opt_t * param) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Xfer(acc:%d, file:%p, buf:%p, len:%llu, %p)\n",
access, file, buffer, length, param);
}
hdfs_options_t * o = (hdfs_options_t*) param;
int xferRetries = 0;
long long remaining = (long long)length;
char* ptr = (char *)buffer;
long long rc;
hdfsFS hdfs_fs = o->fs; /* (void*) */
hdfsFile hdfs_file = (hdfsFile)file; /* (void*) */
while ( remaining > 0 ) {
/* write/read file */
if (access == WRITE) { /* WRITE */
if (verbose >= VERBOSE_4) {
fprintf( stdout, "task %d writing to offset %lld\n",
rank,
offset + length - remaining);
}
if (verbose >= VERBOSE_4) {
printf("\thdfsWrite( %p, %p, %p, %lld)\n",
hdfs_fs, hdfs_file, ptr, remaining ); /* DEBUGGING */
}
rc = hdfsWrite( hdfs_fs, hdfs_file, ptr, remaining );
if ( rc < 0 ) {
ERR( "hdfsWrite() failed" );
}
offset += rc;
if ( hints->fsyncPerWrite == TRUE ) {
HDFS_Fsync( file, param );
}
}
else { /* READ or CHECK */
if (verbose >= VERBOSE_4) {
fprintf( stdout, "task %d reading from offset %lld\n",
rank, offset + length - remaining );
}
if (verbose >= VERBOSE_4) {
printf("\thdfsRead( %p, %p, %p, %lld)\n",
hdfs_fs, hdfs_file, ptr, remaining ); /* DEBUGGING */
}
rc = hdfsPread(hdfs_fs, hdfs_file, offset, ptr, remaining);
if ( rc == 0 ) {
ERR( "hdfs_read() returned EOF prematurely" );
}
if ( rc < 0 ) {
ERR( "hdfs_read() failed" );
}
offset += rc;
}
if ( rc < remaining ) {
fprintf(stdout, "WARNING: Task %d, partial %s, %lld of %lld bytes at offset %lld\n",
rank,
access == WRITE ? "hdfsWrite()" : "hdfs_read()",
rc, remaining,
offset + length - remaining );
if ( hints->singleXferAttempt == TRUE ) {
MPI_CHECK( MPI_Abort( MPI_COMM_WORLD, -1 ), "barrier error" );
}
if ( xferRetries > MAX_RETRY ) {
ERR( "too many retries -- aborting" );
}
}
assert( rc >= 0 );
assert( rc <= remaining );
remaining -= rc;
ptr += rc;
xferRetries++;
}
if(access == WRITE){
// flush user buffer, this makes the write visible to readers
// it is the expected semantics of read/writes
rc = hdfsHFlush(hdfs_fs, hdfs_file);
if(rc != 0){
WARN("Error during flush");
}
}
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Xfer\n");
}
return ( length );
}
/*
* Perform hdfs_sync().
*/
static void HDFS_Fsync(aiori_fd_t * fd, aiori_mod_opt_t * param) {
hdfs_options_t * o = (hdfs_options_t*) param;
hdfsFS hdfs_fs = o->fs; /* (void *) */
hdfsFile hdfs_file = (hdfsFile)fd; /* (void *) */
if (verbose >= VERBOSE_4) {
printf("\thdfsFlush(%p, %p)\n", hdfs_fs, hdfs_file);
}
if ( hdfsHSync( hdfs_fs, hdfs_file ) != 0 ) {
// Hsync is implemented to flush out data with newer Hadoop versions
EWARN( "hdfsFlush() failed" );
}
}
/*
* Close a file through the HDFS interface.
*/
static void HDFS_Close(aiori_fd_t * fd, aiori_mod_opt_t * param) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Close\n");
}
hdfs_options_t * o = (hdfs_options_t*) param;
hdfsFS hdfs_fs = o->fs; /* (void *) */
hdfsFile hdfs_file = (hdfsFile)fd; /* (void *) */
if ( hdfsCloseFile( hdfs_fs, hdfs_file ) != 0 ) {
ERR( "hdfsCloseFile() failed" );
}
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Close\n");
}
}
/*
* Delete a file through the HDFS interface.
*
* NOTE: The signature for ior_aiori.delete doesn't include a parameter to
* select recursive deletes. We'll assume that that is never needed.
*/
static void HDFS_Delete( char *testFileName, aiori_mod_opt_t * param ) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_Delete\n");
}
hdfs_options_t * o = (hdfs_options_t*) param;
char errmsg[256];
/* initialize file-system handle, if needed */
hdfs_connect(o);
if ( ! o->fs )
ERR( "Can't delete a file without an HDFS connection" );
if ( hdfsDelete( o->fs, testFileName, 0 ) != 0 ) {
sprintf(errmsg, "[RANK %03d]: hdfsDelete() of file \"%s\" failed\n",
rank, testFileName);
EWARN( errmsg );
}
if (verbose >= VERBOSE_4) {
printf("<- HDFS_Delete\n");
}
}
/*
* Use hdfsGetPathInfo() to get info about file?
* Is there an fstat we can use on hdfs?
* Should we just use POSIX fstat?
*/
static IOR_offset_t HDFS_GetFileSize(aiori_mod_opt_t * param,
char * testFileName) {
if (verbose >= VERBOSE_4) {
printf("-> HDFS_GetFileSize(%s)\n", testFileName);
}
hdfs_options_t * o = (hdfs_options_t*) param;
IOR_offset_t aggFileSizeFromStat;
IOR_offset_t tmpMin, tmpMax, tmpSum;
/* make sure file-system is connected */
hdfs_connect( o );
/* file-info struct includes size in bytes */
if (verbose >= VERBOSE_4) {
printf("\thdfsGetPathInfo(%s) ...", testFileName);
fflush(stdout);
}
hdfsFileInfo* info = hdfsGetPathInfo( o->fs, testFileName );
if ( ! info )
ERR( "hdfsGetPathInfo() failed" );
if (verbose >= VERBOSE_4) {
printf("done.\n");fflush(stdout);
}
aggFileSizeFromStat = info->mSize;
if (verbose >= VERBOSE_4) {
printf("<- HDFS_GetFileSize [%llu]\n", aggFileSizeFromStat);
}
return ( aggFileSizeFromStat );
}
Reference in New Issue View Git Blame Copy Permalink