/usr/share/octave/packages/parallel-2.2.0/__bw_scheduler__.m is in octave-parallel 2.2.0-1build1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 | ## Copyright (C) 2009 Olaf Till <i7tiol@t-online.de>
##
## This program is free software; you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 3 of the License, or
## (at your option) any later version.
##
## This program is distributed in the hope that it will be useful,
## but WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this program; if not, see <http://www.gnu.org/licenses/>.
function __bw_scheduler__ (f, argfile)
## internal function of the parallel package, not meant for explicit use
## Normally this function runs in a dedicated process, but if someone
## uses it differently, there could be child-processes before; but
## there must not be any.
while (waitpid (-1) > 0) endwhile
## default directories
data_dir = tilde_expand ("~/bw-data/");
state_dir = tilde_expand ("~/.bw-state/");
## default value of minimal time-interval for saving state
min_save_interv = 10; # seconds
## default timeout for connections
connect_timeout = 30; # seconds
## configuration files
systemrc = fullfile (OCTAVE_HOME (), "share/octave/site/m/startup/bwrc");
userrc = "~/.bwrc";
## Read system-wide configuration, should provide
## "computing_machines", a cell-array with addresses, and
## "central_machine", a single address.
source (systemrc);
## Read per-user configuration.
if (! isempty (stat (userrc)))
source (userrc);
endif
## filenames
stfn = fullfile (state_dir, sprintf ("%s-%s.state", f, argfile));
pidfn = fullfile (state_dir, sprintf ("%s-%s.pid", f, argfile));
lfn = fullfile (state_dir, sprintf ("%s-%s.lock", f, argfile));
## read arguments
try
args = __bw_load_variable__ (fullfile (data_dir, argfile));
catch
state.scheduler_msg = lasterr;
__bw_secure_save__ (stfn, state, "state");
return;
end_try_catch
if (! iscell (args) || (rows (args) > 1 && columns (args) > 1))
state.scheduler_msg = "arguments no one-dimensional array\n";
__bw_secure_save__ (stfn, state, "state");
return;
endif
n_args = length (args);
## racing condition, but might avoid some blocked schedulers
if (__bw_is_locked__ (lfn))
return;
endif
## assure unique access to state-file with an advisory locked lockfile
lfd = __bw_lock_file__ (lfn);
## note pid in an extra file
if ((pidfd = fopen (pidfn, "w")) < 0)
return;
endif
fprintf (pidfd, "%i", getpid ());
fclose (pidfd);
## initialize state
if (isempty (stat (stfn)))
state.results = state.msg = cell (n_args, 1);
state.active = state.ready = state.error = zeros (n_args, 1);
state.scheduler_msg = "";
save ("-binary", stfn, "state");
else
state = __bw_load_variable__ (stfn);
if (all (state.ready))
## nothing to do, so quit
return;
endif
state.scheduler_msg = "";
## reset argument information for a running scheduler
state.active = zeros (n_args, 1);
endif
if (! exist ("computing_machines") || isempty (computing_machines))
state.scheduler_msg = "no computing machines given\n";
__bw_secure_save__ (stfn, state, "state");
return;
endif
## initialize machine information
m_n = length (computing_machines);
m_njobs = m_irresponsive = m_active = pids = pipesr = ...
pipesw = m_just_tried = zeros (m_n, 1);
## fork one permanent subprocess per machine
cmd = "ssh";
ssh_timeout_opt_str = sprintf ("-o ConnectTimeout=%i", ...
max (round (connect_timeout), 0));
for id = 1:m_n
[pdrp, pdwc, err, msg] = pipe ();
if (err)
state.scheduler_msg = ...
sprintf ("error %i in creating pipe, msg: %s\n", err, msg);
__bw_secure_save__ (stfn, state, "state");
return;
endif
[pdrc, pdwp, err, msg] = pipe ();
if (err)
state.scheduler_msg = ...
sprintf ("error %i in creating pipe, msg: %s\n", err, msg);
__bw_secure_save__ (stfn, state, "state");
return;
endif
[pids(id), msg] = fork ();
if (pids(id) == 0) # child process
pclose (pdrp);
pclose (pdwp);
fcntl (pdrc, F_SETFL, O_SYNC);
while (true) # no break, process killed from parent process
cmd_args = {"-o PasswordAuthentication=no", ...
ssh_timeout_opt_str, ...
computing_machines{id}, ...
"octave", ...
"-q", ...
"--eval", ...
(sprintf ...
("\"__bw_computing_machine__ (\\\"%s\\\")\"", ...
f)), ...
"2>/dev/null"};
constart = time;
[pdw, pdr, pid] = popen2 (cmd, cmd_args, 1);
while (true) # break if eof on pdr
arg_id = __bw_prcv__ (pdrc).psend_var;
__bw_psend__ (pdw, args{arg_id});
__bw_psend__ (pdw, arg_id);
fflush (pdw);
try
if (ismatrix (tp = __bw_prcv__ (pdr)))
break;
endif
catch
break;
end_try_catch
tp = tp.psend_var;
if (tp == 2) # computing function returned error
try
if (ismatrix (tp = __bw_prcv__ (pdr)))
break;
endif
catch
break;
end_try_catch
tp = tp.psend_var;
__bw_psend__ (pdwc, 2);
__bw_psend__ (pdwc, tp);
fflush (pdwc);
else # success
try
if (ismatrix (tp = __bw_prcv__ (pdr)))
break;
endif
catch
break;
end_try_catch
tp = tp.psend_var;
__bw_psend__ (pdwc, 0);
__bw_psend__ (pdwc, tp);
fflush (pdwc);
endif
endwhile
waitpid (pid);
pclose (pdr);
pclose (pdw);
__bw_psend__ (pdwc, 1); # computing machine (got) unreachable
fflush (pdwc);
if ((rest = connect_timeout + constart - time) > 0)
pause (rest);
endif
endwhile
## not reached, but left here lest they are needed sometime
pclose (pdrc);
pclose (pdwc);
__exit__ (0);
elseif (pids(id) > 0) # parent process
pclose (pdwc);
pclose (pdrc);
pipesr(id) = pdrp;
pipesw(id) = pdwp;
fcntl (pipesr(id), F_SETFL, O_SYNC);
else # error
state.scheduler_msg = ...
sprintf ("error %i in forking, msg: %s\n", pids(id), msg);
__bw_secure_save__ (stfn, state, "state");
return;
endif
endfor
## scheduling section
last_saved = time;
while (! all (state.ready))
args_unused = find (! (state.ready | state.active));
m_free = cat (1, ...
find (! (m_active | m_irresponsive)), ...
find ((! (m_active | m_just_tried)) & ...
m_irresponsive), ...
find ((! m_active) & m_just_tried & m_irresponsive));
## there should always be free childs here, give them a task
for id = 1:min (length (m_free), length (args_unused))
## tell child to use this argument
__bw_psend__ (pipesw(m_free(id)), args_unused(id));
fflush (pipesw(m_free(id)));
## mark child active (busy) and note argument of machine
m_active(m_free(id)) = args_unused(id);
## mark argument active
state.active(args_unused(id)) = true;
endfor
[np, act_idx] = select (pipesr, [], [], -1);
## respond to childs answers
for id = act_idx
try
res = __bw_prcv__ (pipesr(id)).psend_var;
catch
state.scheduler_msg = ...
sprintf ("error in reading state from child %i\n", id);
__bw_secure_save__ (stfn, state, "state");
return;
end_try_catch
switch res
case 0 # success
m_irresponsive(id) = false;
m_just_tried(id) = false;
m_njobs(id)++;
state.active(m_active(id)) = false;
state.ready(m_active(id)) = true;
try
state.results{m_active(id)} = __bw_prcv__ (pipesr(id)).psend_var;
catch
state.scheduler_msg = ...
sprintf ("error in reading result %i from child %i\n", ...
m_active(id), id);
__bw_secure_save__ (stfn, state, "state");
return;
end_try_catch
m_active(id) = 0;
case 1 # computing machine (got) unreachable
state.active(m_active(id)) = false;
m_irresponsive(id) = true;
m_just_tried(id) = true;
m_active(id) = 0;
case 2 # computing function returned error
m_irresponsive(id) = false;
m_just_tried(id) = false;
m_njobs(id)++;
state.active(m_active(id)) = false;
state.ready(m_active(id)) = true;
state.error(m_active(id)) = true;
try
state.msg{m_active(id)} = __bw_prcv__ (pipesr(id)).psend_var;
catch
state.scheduler_msg = ...
sprintf ("error in reading functions error message for args %i from child %i\n", ...
m_active(id), id);
__bw_secure_save__ (stfn, state, "state");
return;
end_try_catch
m_active(id) = 0;
otherwise # invalid
state.scheduler_msg = ...
sprintf ("child %i returned invalid state\n", res);
__bw_secure_save__ (stfn, state, "state");
return;
endswitch
endfor
## update statefile, but not if last update was a short time ago
state.machines.active = m_active;
state.machines.irresponsive = m_irresponsive;
state.machines.njobs = m_njobs;
if ((tp = time) - last_saved >= min_save_interv)
__bw_secure_save__ (stfn, state, "state");
last_saved = tp;
endif
if (all ((m_just_tried | ! m_irresponsive)(! m_active)))
m_just_tried = zeros (m_n, 1);
endif
endwhile
## final update of statefile
__bw_secure_save__ (stfn, state, "state");
## terminate child processes, no coredump
for pid = pids'
system (sprintf ("kill -9 %i", pid));
waitpid (pid);
endfor
## release lockfile
__bw_unlock_file__ (lfd);
## if all is ready, unlink lockfile; earlier unlinking breaks locking
unlink (lfn);
## unlink pidfile
unlink (pidfn);
endfunction
|