How to install a BOINC server to test SpeQuloS
INRIA - EDGI JRA2


System used : Debian 6.0 "Squeeze"


  New BOINC system
  ================

To start installing a new BOINC system from scratch, you can do the
following:

Prepare the system:

  export BOINC_ADM=boincadm
  export DB_NAME=boinc_stp_db
  export DB_USER=$BOINC_ADM
  export DB_PASS=edgi
  export WWW_HOST=http://boinc-server.org
  export PROJECT_NAME=boinc-stp
  export PROJECT_DIR=/opt/$PROJECT_NAME
  export SPEQULOS_PLUGIN_DIR=~/spequlos/DG-plugins/BOINC
  
  apt-get update
  apt-get install subversion build-essential apache2 php5 mysql-server php5-gd php5-cli php5-mysql python-mysqldb libtool automake autoconf pkg-config libmysql++-dev libssl-dev
  addgroup --system  $BOINC_ADM
  adduser www-data $BOINC_ADM

Download the BOINC server

  svn co http://boinc.berkeley.edu/svn/branches/server_stable boinc
  cd boinc

Ensure that sched/target_batch.cpp exists, as this is the component that
SpeQuloS uses.

Increase max assignment table size as needed:

- Increase MAX_ASSIGNMENTS in sched/sched_shmem.h
- Set the corresponding variable MAX_TO_ASSIGN in the file
sched/target_batch.cpp to the same value.

Set dedicated (cloud) user id in sched/target_batch.cpp via the
DEDICATED_USER_ID constant.

Build the BOINC server

  ./_autosetup 
  ./configure --disable-client
  make

Prepare the database

  mysql -u root -p mysql -e "DROP DATABASE IF EXISTS $DB_NAME; \
   CREATE DATABASE IF NOT EXISTS $DB_NAME; \
   GRANT ALL PRIVILEGES ON $DB_NAME.* TO $DB_USER@'localhost' IDENTIFIED BY '$DB_PASS';"

Create a project

  cd tools/
  ./make_project --url_base "$WWW_HOST" --db_name "$DB_NAME" --db_user $DB_USER  --db_passwd "$DB_PASS" --drop_db_first --delete_prev_inst --project_root $PROJECT_DIR --test_app "$PROJECT_NAME"

Configure the project

  cd $PROJECT_DIR
  chown root:$BOINC_ADM  -R .
  chmod g+w -R .
  chmod 02770 -R upload html/cache html/inc html/languages html/languages/compiled html/user_profile
  chmod    o+x html/inc
  chmod -R o+r html/inc
  chmod    o+x html/languages/
  chmod    o+x html/languages/compiled
  crontab -e #Add:  0-59/5 * * * * /PATH/TO/PROJECT/DIR/bin/start --cron
  cp $PROJECT_NAME.httpd.conf /etc/apache2/sites-available/
  #vim html/project/project.inc 
  a2ensite $PROJECT_NAME.httpd.conf
  /etc/init.d/apache2 reload

Set the administrator password for the web access:

  htpasswd -c html/ops/.htpasswd $BOINC_ADM

Preconfigure a client that will be used by the SpeQuloS Cloudworkers

- Edit $SPEQULOS_PLUGIN_DIR/create_cloud_id.sql and change the first "1"
to chosen DEDICATED_USER_ID.
- Create the client in the database:
  mysql -u $DB_USER --password="$DB_PASS" $DB_NAME <
$SPEQULOS_PLUGIN_DIR/create_cloud_id.sql


Specify the base project_dir in config.xml, in the <config> section:
    <config>
        ...
        <project_dir>
        /PATH/TO/PROJECT/DIR
        </project_dir>
        ...
    </config>


Install the SpeQuloS BOINC plugins:
  cp $SPEQULOS_PLUGIN_DIR/*.php html/ops/


Enable assignment to SpeQuloS Cloudworkers, by adding the following to
config.xml, in the <config> section:
  <config>
      ...
      <debug_assignment/>
      <enable_assignment/>
      ...
  </config>


Enable exporting of BOINC statistics (used to export credits
information), by adding to config.xml, in the <config> section:
      <tasks>
          <task>
              <cmd>db_dump -d 2 -dump_spec ../db_dump_spec.xml</cmd>
              <output>db_dump.out</output>
              <period>24 hours</period>
          </task>
      </tasks>


Create the initial statistics files:
  bin/db_dump -d 2 -dump_spec ../db_dump_spec.xml


Install the target_batch from the compiled source dir to your project
cgi-bin:

  cp $BOINC_SRC/sched/target_batch $PROJECT_DIR/cgi-bin/

An input parameter of the form batch=[id] can be passed to target_batch
via url or commmand line. See more details in target_batch.cpp.
Depending on your web server settings, you may have to inform your web
server to allow target_batch be called externally as a cgi script.

Install the example application

  bin/xadd 
  bin/update_versions 

To disable automatic creation of workunit for the example application,
comment enclose by <!-- and --> the following lines in config.xml:

      <!-- 
      <daemon>
        <cmd>
          sample_work_generator -d 3
        </cmd>
      </daemon>
      -->

Start the server

  bin/start 

To configure the properties of the workunit for the example application,
edit the file templates/uc_wu, i.e.:

  <file_info>
      <number>0</number>
  </file_info>
  <workunit>
      <file_ref>
          <file_number>0</file_number>
          <open_name>in</open_name>
      </file_ref>
      <rsc_memory_bound>1e8</rsc_memory_bound>
      <command_line>-cpu_time 10</command_line>
  </workunit>

You can submit some workunits by using (from within the project
directory):

  $SPEQULOS_PLUGIN_DIR/test/create_test_work.pl

Beforehand, you must edit the DEDICATED_USER_ID variable in this file
according to your project parameters.

Then, for example, to create a batch of low-priority tasks with the
BOINC default example_app, run:

  ./create_test_work.pl -appname example_app -numwu 10 -batch 0 -label "lowpri" -db $DB_NAMEs -password $PASSWORD -wu "templates/example_app_in.xml" -result "templates/example_app_out.xml"


  Existing BOINC system
  =====================

If you have an existing system, you can simple update your BOINC source
from svn to revision: 25582 or later.

You will get a file sched/target_batch.cpp, which is used by SpeQuloS to
target a batch of uncompleted wus to dedicated hosts.

See the documentation above to :

- increase the assignment table
- configure the DEDICATED_USER_ID
- enable the assignment feature in your project configuration file
- copy the target_batch file to cgi-bin directory.

Then, you can upgrade your project via:
cd $BOINC_SRC/source/tools
./upgrade $PROJECT_NAME