Changeset - 355de37ac5f1
[Not reviewed]
default
0 2 0
ethanzonca - 16 years ago 2009-09-27 19:33:39

Minor edits
2 files changed with 15 insertions and 29 deletions:
0 comments (0 inline, 0 general)
TODO
Show inline comments
 
Key:
 
*-An important Item that needs to be finished
 
B-Initial of person who needs to fix this (Binki)
 
M-LordofWar needs to fix this
 
E-normaldotcom needs to fix this
 
x-finished task
 
=-Long-term goal
 

	
 
Build
 

	
 
Master
 
*-Code dummy data/processes to test the current queueing system / watchdog / etc.
 
*-Add calls to remotio
 
 -Write more meat into the pseudo-code main() to pull everything together
 
 -Write a stub for getting info from the tarball/validifying the tarball. Read distren-job.xml, a file in the tarball, to find out 1. which rendering system to use (that system, e.g. blender/povray, can read more specifics, such as name of file to pass to blender and frames. Options common between different systems will be handled in common as best as possible)
 
 -Write a stub for publishing file and constructing job description so that the job can be shared
 
*-Other XML-writing/reading such as current stats / jobs / data / etc, xml writing for job creation, ...
 
*-charsets: the program should operate in UTF-8. This is partially required by libxml2's (and XML's) usage of UTF-8. We want the program's internal charset to be UTF-8 and figure out if we need w_char/whatever to fulfill UTF-8. I don't believe we do... mabye we just need a declaration that al data stored must be in UTF-8 format.
 
 -rename some struct members and variables to be more program-agnostic and more sensible (e.g., no blender-specific stuff)
 

	
 
Slave
 
 -Fix current bugs
 
*-Add calls to remoteio, once it works or even before it works
 
*-Other XML-writing/reading
 
	+Update exec_blender() to use some struct info, maybe. The struct is being passed to it.
 
*-Other XML-writing/reading: Update exec_blender() to use some struct info, maybe. The struct is being passed to it.
 
=-Make code more flexible for different types of jobs and operating systems (such as paths, libs, etc)
 
 -Add code to allow pausing/resuming of the slave (including a running blender process)
 
 -Add a threads variable for software that doesn't support threads (blender takes care of this by itself)
 
*-Fix line 194 (call to myjob struct that fails miserably)
 
B-Add tmpdir variable (just like the datadir variable) that is compile-time or whatever
 
*-Add code to delete old jobfiles (maybe based on timestamp, etc... make sure they're uploaded first)
 
 -Upload while rendering? Would be nice.
 
*-Finish stub to delete old job data and output
 
=-Upload while rendering
 

	
 
Options
 
 -Rewrite some stuff, try to make it simpler
 
x-Move server and client confuse code into their individual files, rather than in the common file. -- note: there wasn't any client code in options.c. There was the ``server'' section. That section is passed multiple times and provides information to remoteio on how to connect to servers
 
 -Review all confuse interfacing
 
 -Push patch for relative includes in confuse to the confuse developers
 
B-Expand execio() to support stderr
 
 
 
\ No newline at end of file
src/server/distrend.c
Show inline comments
 
@@ -15,362 +15,352 @@
 

	
 
  You should have received a copy of the GNU Affero General Public License
 
  along with DistRen.  If not, see <http://www.gnu.org/licenses/>.
 

	
 
*/
 

	
 
/* This file contains the code which both processes (renders) jobs as a slave, and the code which distributes frames to slaves after receiving them from the client portion of the codebase. */
 

	
 
#include "execio.h"
 
#include "options.h"
 
#include "distrenjob.h"
 
#include "protocol.h"
 
#include "slavefuncs.h"
 
#include "asprintf.h"
 

	
 
#include <confuse.h>
 
#include <stdio.h>
 
#include <stdlib.h>
 
#include <malloc.h>
 
#include <unistd.h> /* getopt */
 
#include <time.h>
 
#include <sys/stat.h>
 
#include <string.h>
 

	
 
/* ******************* Structs ************************ */
 

	
 
// Gets config info from confs
 
struct distrend_config
 
{
 
  cfg_t *mycfg;
 
  struct options_common *options;
 
  struct distrend_listen **listens; /*< Null terminated array of structs */
 
  char *datadir;
 
};
 

	
 
struct general_info {
 
  short int jobs_in_queue; //
 
  unsigned short int free_clients;
 
  unsigned short int rendering_clients;//
 
  unsigned short int total_finished_jobs; //
 
  unsigned int total_frames_rendered; //
 
  unsigned int highest_jobnum;
 
  short int hibernate;
 
} general_info;
 

	
 

	
 

	
 
/*
 
  internally defined funcs's prototypes
 
*/
 
  internally defined funcs's prototypes @TODO: Make all functions nice and proper */
 
void distrenjob_remove(struct distrenjob *head, struct distrenjob *bj);
 

	
 
struct distrenjob *distrenjob_get(struct distrenjob *head, jobnum_t jobnum);
 
void distrenjob_enqueue(struct distrenjob *head, struct distrenjob *job);
 
void mortition(struct distrenjob *head, struct distrenjob *job);
 

	
 
/* Global Vars, try to cut down on these */
 
/* Global Vars, eliminate these */
 
jobnum_t jobnum = 0; // The next job number to create in the queue
 
int hcfjob; // Highest consecutively finished job
 
int highest_jobnum; // The job number of the most recently created job, this is used when creating new jobs
 

	
 

	
 
/* ********************** Functions ************************* */
 

	
 
/** Dumps all data in RAM to an xml file (such as current jobs, etc) which is parsed by start_data. Remember to invoke this before shutting down! */
 
int xml_dump()
 
{
 
  return 0;
 
}
 
/**
 
   Performs command stored in a client's request.
 
*/
 
int distrend_do()
 
{
 
  return 0;
 
}
 
/**
 
   Accepts a client's connection
 
 */
 
void distrend_accept()
 
{
 

	
 
}
 
/**
 
   Frees the action
 
*/
 
void distrend_action_free()
 
{
 

	
 
}
 
/**
 
   Start listening
 
*/
 
void distrend_listen()
 
{
 

	
 
}
 
/**
 
   Stop listening
 
*/
 
void distrend_unlisten()
 
{
 

	
 
}
 
/**
 
   This is probably just a placeholder for remotio
 
*/
 
void remotio_send_to_client()
 
{
 
	// I am futile!
 
}
 

	
 
/** Fill variables after crash / shutdown from XML dumps */
 
int start_data()
 
{
 
  general_info.hibernate = 0;
 
  struct stat buffer;
 
  if(stat(SYSCONFDIR "/data.xml", &buffer) == 0){
 

	
 
    // @TODO: retrieve total_finished_jobs and total_finished_frames from xml file
 

	
 
    fprintf(stderr,"Parsing XML files and restoring previous state...\n");
 
    return 1;
 
  }
 
  else{
 
    general_info.total_finished_jobs = 0;
 
    general_info.total_frames_rendered = 0;
 
    fprintf(stderr,"Can't find XML dump, starting up fresh.\n");
 
    return 2;
 
  }
 
  free(&buffer); // @TODO: Is this pointless?
 
}
 

	
 
/** Finish-Setter: Sets a frame to the "completed" status.*/
 
void finish_frame(struct distrenjob *head, struct distrenjob *distrenjob, int frame)
 
{
 
  distrenjob->frameset[frame].status = FRAMESETSTATUS_DONE;
 
  distrenjob->total_render_time = distrenjob->total_render_time + (clock() - distrenjob[jobnum].frameset[frame].start_time);
 
  distrenjob->completed_frames++;
 
  distrenjob->assigned_frames--;
 
  general_info.total_frames_rendered++; // Increase total frames var for stats
 

	
 
  if(distrenjob->completed_frames == distrenjob->total_frames)
 
    {
 
      mortition(head, distrenjob);
 
    }
 
}
 

	
 
/** mortition check to see if a job is actually done by scanning the folder of the job to make sure all frames are present*/
 
// called mortition because it checks the finished "dead" job
 
/** "mortition" check to see if a job is actually done by scanning the folder of the job to make sure all frames are present*/
 
void mortition(struct distrenjob *head, struct distrenjob *job)
 
{
 
  short int isJobDone;
 
  int counter;
 
  char *path_and_number;
 
  struct stat buffer;
 

	
 
  isJobDone = 1; // set isJobDone to true
 
  for(counter = 0; counter < job->total_frames; counter++)
 
    {
 
      _distren_asprintf(&path_and_number, "stor/job%d/out/%d.%s", job->jobnum, job->frameset[counter].num, job->output_format);
 
      if(stat(path_and_number, &buffer) != -1)
 
        {
 
          job->frameset[counter].status = FRAMESETSTATUS_UNASSIGNED;
 
          job->completed_frames--;
 
          isJobDone = 0; // if a missing frame is found, set isJobDone to false
 
        }
 
    }
 

	
 
  if(isJobDone) // if all frames were accounted for
 
    {
 
      distrenjob_remove(head, job);
 
      distrenjob_free(&job);
 
      general_info.jobs_in_queue--;
 
    }
 
  else{
 
    job->prev_frame_index = 0; // if the job isn't done, have frame_finder() start from the first frame, allowing it to see the frames that are now unassigned
 
  }
 
}
 

	
 
/**
 
   creates a structure from starting data, then calls another function to actually add that struct to the queue
 
*/
 
/** creates a structure from starting data, then calls another function to actually add that struct to the queue */
 
void prepare_distrenjob(struct distrenjob *head, int type, char *name, char *submitter, char *email, int priority, int start_frame, int end_frame)
 
{
 
  int counter2;
 
  int counter;
 

	
 
  struct distrenjob *distrenjob;
 
  distrenjob_new(&distrenjob);
 

	
 
  distrenjob->type = 1;
 
  distrenjob->name = name;
 
  distrenjob->submitter = submitter;
 
  distrenjob->email = email;
 
  distrenjob->priority = priority;
 
  distrenjob->total_frames = (end_frame - start_frame + 1); // sets the total number of frames in animation for status purposes
 

	
 
  /* prepares all the frames by setting that status to "unassigned" */
 
  counter2 = start_frame;
 
  for(counter = 0; counter <= (end_frame- start_frame + 1); counter++){
 
    distrenjob->frameset[counter].num = counter2;
 
    distrenjob->frameset[counter].status = FRAMESETSTATUS_UNASSIGNED;
 

	
 
    counter2++;
 
  }
 

	
 
  /* add job to queue */
 
  distrenjob_enqueue(head, distrenjob);
 

	
 
  general_info.jobs_in_queue++;
 
}
 

	
 

	
 
/* distrenjob_enqueue: This function adds the job to the queue based on its priority */
 
/** distrenjob_enqueue: This function adds the job to the queue based on its priority */
 
void distrenjob_enqueue(struct distrenjob *head, struct distrenjob *job) {
 
  struct distrenjob *prev_job = head; // create pointer to previous job
 
  struct distrenjob *current_job;     // create pointer to current_job (which job is being compared to)
 

	
 
  // iterate through linked list of jobs
 
  for(current_job = head; current_job != NULL; current_job = current_job->next){
 
    if(current_job == NULL){ // if it has reached the end of the list, add job there
 
      current_job = job;
 
      break;
 
    }
 
    else if(job->priority < current_job->priority){ // if job's priority is less than current_job's priority, insert job
 
      prev_job->next = job;                        // keep in mind 1 is the highest priority given to jobs, head has a
 
      job->next = current_job;                     // priority of zero so it will always be before other jobs
 
      break;
 
    }
 

	
 
    prev_job = current_job;
 
  } /* for(current_job) */
 
}
 

	
 
/**
 
   @arg head I may end up changing the head if job == head
 
 */
 
/** Changes the priority of an existing (and maybe running) job. @arg head I may end up changing the head if job == head */
 
void change_job_priority(struct distrenjob *head, struct distrenjob *job, int new_priority){
 
  distrenjob_remove(head, job);
 
  job->priority = new_priority;
 
  struct distrenjob *current_job;
 
  struct distrenjob *prev_job = head;
 

	
 
  if(job->frameset[0].status == FRAMESETSTATUS_UNASSIGNED)
 
    /* if job was not yet started */
 
    distrenjob_enqueue(head, job);
 
  else{ // if job has already been started then place it before the jobs with the same priority
 
  else{ // if job has already been started, place it before the jobs with the same priority
 
    // iterate through linked list of jobs
 
    for(current_job = head; current_job != NULL; current_job = current_job->next){
 
      if(current_job == NULL){ // if it has reached the end of the list, add job there
 
        current_job = job;
 
        break;
 
      }
 
      else if(job->priority <= current_job->priority){ // if job's priority is less than or equal to current_job's priority, insert job
 
        prev_job->next = job;                        // keep in mind 1 is the highest priority given to jobs, head has a
 
        job->next = current_job;                     // priority of zero so it will always be before other jobs
 
        break;
 
      }
 

	
 
      prev_job = current_job;
 
    }
 
  }
 
}
 

	
 
/**
 
  Frame Finder: matches your computer up with a lovely frame to render
 
  TODO: Major issue here, the client needs to know the frame number, AND the job number!
 
  Notice that this function starts by looking at the oldest job first
 

	
 
	TODO: Link this up with the main() function to check if there are frames available or not and provide jobnum/framenum to the client
 

	
 
  Frame Finder: matches your computer up with a lovely frame to render, starts looking at oldest job first
 
  @TODO: We must return both jobnum and framenum
 
  @TODO: Add calls in main()
 
  @return 0 success, other: error
 
*/
 
int find_jobframe(struct distrenjob *head, struct distrenjob **job, struct frameset **frame)
 
{
 
  if(general_info.hibernate)
 
    return 1;
 

	
 
  unsigned int frame_counter;
 
  short int your_job_type;
 
  unsigned short int found;
 

	
 
  struct distrenjob *distrenjob_ptr;
 

	
 
  your_job_type = 0;
 
  found = 0;
 
  /* iterate through jobs from first to last */
 
  for(distrenjob_ptr = head->next; (!found && !distrenjob_ptr) && !distrenjob_ptr->hibernate; distrenjob_ptr = distrenjob_ptr->next)
 
    for(frame_counter = (distrenjob_ptr->prev_frame_index + 1); !found && frame_counter < distrenjob_ptr->total_frames; frame_counter ++)
 
      if(distrenjob_ptr->frameset[frame_counter].status == FRAMESETSTATUS_UNASSIGNED) // jobframe found
 
	{
 
	  found = 1;
 
	  your_job_type = distrenjob_ptr->type;
 
	  distrenjob_ptr->frameset[frame_counter].status = FRAMESETSTATUS_ASSIGNED;
 
	  distrenjob_ptr->frameset[frame_counter].start_time = clock();
 
	  distrenjob_ptr->assigned_frames++;
 
	  distrenjob_ptr->prev_frame_index = frame_counter;
 
	}
 

	
 
  if(!found)
 
    {
 
      fprintf(stderr, "No more jobs to render\n");
 
      sleep(1); /*< @todo eliminate the need for this line*/
 
      return 1;
 
    }
 

	
 
  *job = distrenjob_ptr;
 
  *frame = &distrenjob_ptr->frameset[frame_counter];
 

	
 
  return 0;
 
}
 

	
 
/** Checks for dead, laggy, or stale slaves */
 
/** Checks for dead, latent, or stale slaves */
 
void frame_watchdog(struct distrenjob *distrenjob_head)
 
{
 
  struct distrenjob *distrenjob_ptr;
 
  unsigned int counter;
 

	
 
  /*watchdog_forgiveness = seconds of forgiveness before frame is re-assigned */
 
  distrenjob_ptr = distrenjob_head;
 

	
 
  for(distrenjob_ptr = distrenjob_head; distrenjob_ptr; distrenjob_ptr = distrenjob_ptr->next)
 
  /* iterate through jobs */
 

	
 
    /* if the job has been started, checks by seeing if either to first or second frame has been started */
 
    if(distrenjob_ptr->frameset[0].status != FRAMESETSTATUS_UNASSIGNED || distrenjob_ptr->frameset[1].status != FRAMESETSTATUS_UNASSIGNED)
 
      for(counter = 0; counter < distrenjob_ptr->total_frames; counter ++)
 
      /* iterate through all frames for this job*/
 

	
 
        if((distrenjob_ptr->frameset[counter].start_time + distrenjob_ptr->watchdog_forgiveness) < clock())
 
        /*
 
        If frame is not completed within the number of seconds specified by watchdog_forgiveness
 
        Then change the frame status to unassigned
 
        */
 
          distrenjob_ptr->frameset[counter].status = FRAMESETSTATUS_UNASSIGNED;
 
}
 

	
 
/**
 
   Finds a distrenjob struct based on the jobnum
 
   @arg jobnum job number to search for
 
   @return NULL on job doesn't exist
 
 */
 
struct distrenjob *distrenjob_get(struct distrenjob *head, jobnum_t jobnum)
 
{
 
  struct distrenjob *distrenjob_ptr;
 

	
 
  /*
 
    The conditions of the for loop will leave distrenjob_ptr at NULL if the end of the list is reached. It will leave it pointing to the correct job if it is found.
 
   */
 
  for(distrenjob_ptr = head;
 
      distrenjob_ptr
 
        && distrenjob_ptr->jobnum != jobnum;
 
      distrenjob_ptr = distrenjob_ptr->next);
 

	
 
  return distrenjob_ptr;
 
}
 

	
 

	
 
/**
 
   Removes a distrenjob from the distrenjob linked list. It does not free the distrenjob, however. You should do that with distrenjob_free() from distrenjob.h
 

	
 
@@ -459,65 +449,64 @@ int main(int argc, char *argv[])
 
  struct distrenjob head;
 
  head.priority = 0; // make head have the highest priority
 

	
 
  int cont;
 
  struct distrend_listenset *listenset;
 
  struct distrend_config *config;
 

	
 
  enum clientstatus
 
  {
 
    CLIENTSTATUS_UNINITIALIZED = 0,
 
    CLIENTSTATUS_BUSY = 1,
 
    CLIENTSTATUS_IDLE = 2
 
  } clientstatus;
 

	
 
  cont = 1;
 
  memset(&head, '\0', sizeof(struct distrenjob));
 

	
 
  start_data(); // Starts fresh or loads data from xml dump. Should we grab the return?
 

	
 
  distrend_do_config(argc, argv, &config);
 

	
 
  distrend_listen(&listenset, config);
 
  /* This is called the "main loop" */
 
  while(cont)
 
    {
 
      struct distren_action *action;
 
      int clientsays = 0; /*< temporary example variable, will be replaced when we can handle messages */
 

	
 
      distrend_accept(&action);
 
      cont = distrend_do(action);
 

	
 
      /* Make the following code more event-driven */
 
      frame_watchdog(&head);
 

	
 

	
 
      struct frameset *frame;
 
      struct distrenjob *job;
 

	
 
      /* If the client is idle, must be modified for climbing through linked list of clients (client->clientnum) */
 
      if(clientstatus == CLIENTSTATUS_IDLE)
 
	{
 
	  int returnnum = find_jobframe(&head, &job, &frame); // Finds a frame to render
 
	  if(returnnum)
 
	    {
 
	      fprintf(stderr,"No frames are available to render at this time. Idling...\n");
 
	      sleep(10);
 
	    }
 
	  else
 
	    /* returnnum == 0 */
 
	    remotio_send_to_client(frame->num, job->jobnum); // Pseudo-sends data to client
 
	}
 
      /* If the client states that they finished the frame */
 
      	if(clientsays == DISTREN_REQUEST_DONEFRAME){
 
      	  clientstatus = CLIENTSTATUS_IDLE; // Sets the client back to idle
 
      	  finish_frame(&head, job, frame->num); // @TODO: Check that finish_frame really gets the jobnum somehow (Matt: can we give it the pointer to the job?, if not we can use distrenjob_get(&head, jobnum))
 
      	  finish_frame(&head, job, frame->num); // @TODO: Make sure this actually works.
 
      	}
 

	
 
      distrend_action_free(action);
 
    }
 

	
 
  distrend_unlisten(listenset);
 
  distrend_config_free(config);
 

	
 
  return 0;
 
}
0 comments (0 inline, 0 general)