Bash Scripting

welby · 08-12-2003, 11:14 AM

Short version: Anyone know a way to script a file copy so that x amount of data copies to a location, flushes, and the copy resumes to gather x amount of data again?

We're implementing a DVD system to backup fiels on remote servers. The way the DVD system archives data is this:

Data is put into a cache directory of the local machine, then moves the data onto DVD. If the amount of data going into the cache outstips the amount of data going out of the cache, the copy times out. Due to disk space limitations, I've only got about 20GB of cache space available to give the DVD system, but I've got 80+ GB of data on various servers.

The directory structure is too fragmented to parse the directories in a reference file. (18,000 directories on server, for example) I've figured out the while statement (code follows) to check the parameters of the cache, but can't figure a way to limit the initial data copy until the system is done archiving.

Script:

#!/bin/sh

SERVERLIST_FILENAME="serverlist.txt"

DELAY_THRESHOLD=0

DELAY_TIME="60s"

DELAYED_EVENTS=99999

poll_for_mmparam_delayedevents() {
echo I am polling right now.... I am waiting a really long time...
DELAYED_EVENTS=`mmparam TEST | grep "Number of delayed events:" | awk '{print $5}'`
sleep $DELAY_TIME
}

for BACKUP_PATH in `cat $SERVERLIST_FILENAME | grep -v "^#"`
do
SOURCE_PATH=`echo $BACKUP_PATH | awk -F : '{print $1}'`
DESTINATION_PATH=`echo $BACKUP_PATH | awk -F : '{print $2}'`
echo I am backing up $SOURCE_PATH to $DESTINATION_PATH

cp -RPd $SOURCE_PATH $DESTINATION_PATH

# Wait in loop until delayed events are all processed. Only
# then should we move on to copy the next server.
while [ $DELAYED_EVENTS -gt $DELAY_THRESHOLD ]
do
poll_for_mmparam_delayedevents
done

# reset to avoid a potential race-condition
DELAYED_EVENTS=99999
done

Contents of the serverlist file (typically)

/mnt/SERVER1/data/Images:/BACKUP/mount/SERVER1/full/data/Images
/mnt/SERVER1/data/HOME:/BACKUP/mount/SERVER1/full/data/HOME
/mnt/SERVER1/data/APPS:/BACKUP/mount/SERVER1/full/data/APPS
/mnt/SERVER1/data/ADMIN:/BACKUP/mount/SERVER1/full/data/ADMIN

Ideas are welcomed.

crabboy · 08-12-2003, 04:28 PM

How about a C Program?

I just hacked up this C program that I think will do what you need.

It takes three parameters. The source directory, the destination directory and the size you want to pause at. IMPORTANT: Make sure that the dest directory is not in the source tree.

When the program has copied x amount of data to the destination it will pause until it gets a SIG_HUP. The script that is pulling the data will have to delete the files it has already processed and call "kill -HUP <pid>" to restart the copy.

Source directory should probably be relative (but not required). Destination does not have to be. Try it out on a smaller temporary tree.

Is this even close to what you are looking for?

Code:

#include <stdio.h>
#include <pwd.h>
#include <stdio.h>
#include <fcntl.h>
#include <string.h>
#include <dirent.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/param.h>
#include <signal.h>

#define BUFFER_SIZE 2048

int currentSize = 0;

static void sig_hup()
{
   if ((int) signal( SIGHUP, SIG_IGN) == -1 )
      perror("Signal Error");

   if ((int)signal( SIGHUP, sig_hup) == -1 )
      perror("Signal Error");

   return;
}

int copyfile( char * src, char * dest )
{

   int fdIn = 0;
   int fdOut = 0;
   int iReadSize = 0;
   int iWriteSize = 0;
   char szBuffer[ BUFFER_SIZE + 1 ];

   fdIn = open( src, O_RDONLY );
   fdOut = creat( dest, S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH );

   while ( (iReadSize = read( fdIn, szBuffer, BUFFER_SIZE )) > 0 )
   {
      iWriteSize = write( fdOut, szBuffer, iReadSize );

   } // end while

   close( fdIn );
   close( fdOut );
}

void listdir(char * directory, char * destdir, int maxsize)
{
    struct stat FileInfo;
    char name[MAXPATHLEN];
    struct stat dirinfo;
    struct dirent * files;
    DIR * dirin;
    int len, a, log;
    char newdir[MAXPATHLEN];
    char newfile[MAXPATHLEN];

    if (lstat(directory, &dirinfo) < 0)
    {
        perror (directory);
        return;
    }
    if (!S_ISDIR(dirinfo.st_mode))
        return;


    if ((dirin = opendir(directory)) == NULL)
    {
        perror(directory);
        return;
    }
    while ((files = readdir(dirin)) != NULL)
    {
        sprintf(name, "%s/%s", directory, files -> d_name);
        if (lstat(name, &FileInfo) < 0)
            perror(name);


        if ( S_ISDIR( FileInfo.st_mode ) &&
             strcmp(files -> d_name, ".") != 0 &&
             strcmp(files -> d_name, "..") != 0 )
        {
           sprintf( newdir, "%s/%s/", destdir, name );
           mkdir(  newdir, S_IRWXU | S_IRWXG | S_IRWXO );

           listdir(name, destdir, maxsize);

        }
        else if ( strcmp(files -> d_name, ".") != 0 &&
                  strcmp(files -> d_name, "..") != 0 )
        {
            if ( (FileInfo.st_size + currentSize) > maxsize )
            {
               pause();
               currentSize = 0;
            }
            currentSize += FileInfo.st_size;
            sprintf( newfile, "%s/%s", destdir, name );
            copyfile( name, newfile );

        }

    }
    closedir(dirin);
}


main(int argc, char * argv[])
{

   char szInitialDirectory[MAXPATHLEN];
   char newdir[MAXPATHLEN];

   if ((int)signal(SIGHUP, sig_hup ) == -1 )
      perror("Signal Error");

   strcpy( szInitialDirectory, argv[1] );

   sprintf( newdir, "%s/%s/", argv[2], argv[1] );
   mkdir(  newdir, S_IRWXU | S_IRWXG | S_IRWXO );

   listdir( szInitialDirectory, argv[2], atoi(argv[3]) );

}

welby · 08-14-2003, 10:36 AM

You sir, are the man.

Thanks!

Linh · 05-11-2004, 01:12 PM

Hi crabboy. Thank you for showing me the link.

I put the program into my Linux box as you have showned using the
program name, source dir, dest dir, and the size but it just hang there.

1) What does the program suppose to do with regard to the source
directory, destination directory and size ?

2) Does the size refers to a size of a file, or a size of a directory ?

3) Please show me how to run this program.

root:/home# ./directory_2 /home /home/linh 200

crabboy · 05-11-2004, 01:21 PM

You probably don't want to use this program as-is. It was writtin for a specific task of copying files and waiting for a signal.
You probably just want to use the listdir() method removing the desitination dir from the parameter list and also remove:

Code:

          if ( (FileInfo.st_size + currentSize) > maxsize )
            {
               pause();
               currentSize = 0;
            }

also remove call to copyfile() all the signal stuff and the mkdir call.

Linh · 05-11-2004, 02:04 PM

Hi crabboy. Thank you for your help. I modified the code as you have suggested. I then ran it using ./directory_2 /home/ee. This gives correct result, but when I used ./directory_2 /home or ./directory_2 /home/
it did not list any file but the stuff shown below continuously and then it resulted in a segmentation fault.

name = Flash
name = .
name = ..
name = Flash
name = .
name = ..
name = Flash
name = .
name = ..

===================================

Code:

#include <stdio.h>
#include <pwd.h>
#include <stdio.h>
#include <fcntl.h>
#include <string.h>
#include <dirent.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/param.h>
#include <signal.h>

#define BUFFER_SIZE 2048

int currentSize = 0;


/*******************************************************/

void listdir(char * directory, int maxsize)
{
    struct stat FileInfo;
    char name[MAXPATHLEN];
    struct stat dirinfo;
    struct dirent * files;
    DIR * dirin;
    int len, a, log;
    char newdir[MAXPATHLEN];
    char newfile[MAXPATHLEN];

    if (lstat(directory, &dirinfo) < 0)
    {
        perror (directory);
        return;
    }
    if (!S_ISDIR(dirinfo.st_mode))

    if ((dirin = opendir(directory)) == NULL)
    {
        perror(directory);
        return;
    }
    while ((files = readdir(dirin)) != NULL)
    {
        printf ("name = %s\n", files -> d_name);
        sprintf(name, "%s/%s", directory, files -> d_name);
        if (lstat(name, &FileInfo) < 0)
            perror(name);


        if ( S_ISDIR( FileInfo.st_mode ) &&
             strcmp(files -> d_name, ".") != 0 &&
             strcmp(files -> d_name, "..") != 0 )
        {
           sprintf( newdir, "%s/%s/", name );
           mkdir(  newdir, S_IRWXU | S_IRWXG | S_IRWXO );

           listdir(name, maxsize);

        }
        else if ( strcmp(files -> d_name, ".") != 0 &&
                     strcmp(files -> d_name, "..") != 0 )
        {
          currentSize += FileInfo.st_size;
          sprintf( newfile, "%s", name );
        }

    }
    closedir(dirin);
}

/*******************************************************/

main(int argc, char * argv[])
{

   char szInitialDirectory[MAXPATHLEN];
   char newdir[MAXPATHLEN];

   strcpy( szInitialDirectory, argv[1] );

   sprintf( newdir, "%s/%s/", argv[2], argv[1] );
   mkdir(  newdir, S_IRWXU | S_IRWXG | S_IRWXO );

   listdir( szInitialDirectory, atoi(argv[3]) );

}

crabboy · 05-11-2004, 10:58 PM

Try this:

Code:

#include <stdio.h>
#include <pwd.h>
#include <stdio.h>
#include <fcntl.h>
#include <string.h>
#include <dirent.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/param.h>
#include <signal.h>

#define BUFFER_SIZE 2048

int currentSize = 0;


/*******************************************************/

void listdir(char * directory )
{
    struct stat FileInfo;
    char name[MAXPATHLEN];
    struct stat dirinfo;
    struct dirent * files;
    DIR * dirin;
    int len, a, log;

    if (lstat(directory, &dirinfo) < 0)
    {
        perror (directory);
        return;
    }
    if (!S_ISDIR(dirinfo.st_mode))
    {
        perror(directory);
        return;
    }

    if ((dirin = opendir(directory)) == NULL)
    {
        perror(directory);
        return;
    }
    while ((files = readdir(dirin)) != NULL)
    {
/*
        printf ("name = %s\n", files -> d_name);
*/
        sprintf(name, "%s/%s", directory, files -> d_name);
        if (lstat(name, &FileInfo) < 0)
            perror(name);


        if ( S_ISDIR( FileInfo.st_mode ) &&
             strcmp(files -> d_name, ".") != 0 &&
             strcmp(files -> d_name, "..") != 0 )
        {
           listdir(name );

        }
        else if ( strcmp(files -> d_name, ".") != 0 &&
                     strcmp(files -> d_name, "..") != 0 )
        {
           printf("%s\n", name );
        }

    }
    closedir(dirin);
}

/*******************************************************/

main(int argc, char * argv[])
{

   char szInitialDirectory[MAXPATHLEN];

   strcpy( szInitialDirectory, argv[1] );

   listdir( szInitialDirectory );

}

bigearsbilly · 05-12-2004, 07:47 AM

what about using 'dd' ?

billy

Linh · 05-12-2004, 09:36 AM

Thank you crabboy for your help.

I ran the new code and it works. I have not compared the new code with the old one. Can you quickly tells me what changed you have made ?

crabboy · 05-12-2004, 10:43 AM

The segfault was caused by this code:

Code:

    if (!S_ISDIR(dirinfo.st_mode))

    if ((dirin = opendir(directory)) == NULL)
    {
        perror(directory);
        return;
    }

The opendir was being skipped and the readdir was casuing the segfault.

I also cleaned up some other stuff that didn't have to be there.

Hko · 05-12-2004, 12:13 PM

@crabboy: I was wondering, is there a reason why you did not leave the signal handler empty like this?

Code:

static void sig_hup() { }

(if so, why?)

crabboy · 05-12-2004, 12:33 PM

It could have been left blank, but I usually add the if blocks as standard practice just in case I want to add some code in the signal handler later.

Hko · 05-12-2004, 12:40 PM

OK. Thanks.