Hide keyboard shortcuts

Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

1# This file is part of ctrl_bps. 

2# 

3# Developed for the LSST Data Management System. 

4# This product includes software developed by the LSST Project 

5# (https://www.lsst.org). 

6# See the COPYRIGHT file at the top-level directory of this distribution 

7# for details of code ownership. 

8# 

9# This program is free software: you can redistribute it and/or modify 

10# it under the terms of the GNU General Public License as published by 

11# the Free Software Foundation, either version 3 of the License, or 

12# (at your option) any later version. 

13# 

14# This program is distributed in the hope that it will be useful, 

15# but WITHOUT ANY WARRANTY; without even the implied warranty of 

16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

17# GNU General Public License for more details. 

18# 

19# You should have received a copy of the GNU General Public License 

20# along with this program. If not, see <http://www.gnu.org/licenses/>. 

21 

22"""Driver functions for each subcommand. 

23 

24Driver functions ensure that ensure all setup work is done before running 

25the subcommand method. 

26""" 

27 

28 

29__all__ = [ 

30 "acquire_qgraph_driver", 

31 "cluster_qgraph_driver", 

32 "transform_driver", 

33 "prepare_driver", 

34 "submit_driver", 

35 "report_driver", 

36 "cancel_driver", 

37] 

38 

39 

40import getpass 

41import logging 

42import os 

43import re 

44import shutil 

45 

46 

47from lsst.daf.butler.core.utils import time_this 

48from lsst.obs.base import Instrument 

49 

50from . import BPS_SEARCH_ORDER, BpsConfig 

51from .pre_transform import acquire_quantum_graph, cluster_quanta 

52from .transform import transform 

53from .prepare import prepare 

54from .submit import submit 

55from .cancel import cancel 

56from .report import report 

57 

58 

59_LOG = logging.getLogger(__name__) 

60 

61 

62def _init_submission_driver(config_file, **kwargs): 

63 """Initialize runtime environment. 

64 

65 Parameters 

66 ---------- 

67 config_file : `str` 

68 Name of the configuration file. 

69 

70 Returns 

71 ------- 

72 config : `lsst.ctrl.bps.BpsConfig` 

73 Batch Processing Service configuration. 

74 """ 

75 config = BpsConfig(config_file, BPS_SEARCH_ORDER) 

76 

77 # Override config with command-line values 

78 # Handle diffs between pipetask argument names vs bps yaml 

79 translation = {"input": "inCollection", 

80 "output_run": "outCollection", 

81 "qgraph": "qgraphFile", 

82 "pipeline": "pipelineYaml"} 

83 for key, value in kwargs.items(): 

84 # Don't want to override config with None or empty string values. 

85 if value: 

86 # pipetask argument parser converts some values to list, 

87 # but bps will want string. 

88 if not isinstance(value, str): 

89 value = ",".join(value) 

90 new_key = translation.get(key, re.sub(r"_(\S)", lambda match: match.group(1).upper(), key)) 

91 config[f".bps_cmdline.{new_key}"] = value 

92 

93 # Set some initial values 

94 config[".bps_defined.timestamp"] = Instrument.makeCollectionTimestamp() 

95 if "operator" not in config: 

96 config[".bps_defined.operator"] = getpass.getuser() 

97 

98 if "uniqProcName" not in config: 

99 config[".bps_defined.uniqProcName"] = config["outCollection"].replace("/", "_") 

100 

101 # make submit directory to contain all outputs 

102 submit_path = config["submitPath"] 

103 os.makedirs(submit_path, exist_ok=True) 

104 config[".bps_defined.submitPath"] = submit_path 

105 

106 # save copy of configs (orig and expanded config) 

107 shutil.copy2(config_file, submit_path) 

108 with open(f"{submit_path}/{config['uniqProcName']}_config.yaml", "w") as fh: 

109 config.dump(fh) 

110 

111 return config 

112 

113 

114def acquire_qgraph_driver(config_file, **kwargs): 

115 """Read a quantum graph from a file or create one from pipeline definition. 

116 

117 Parameters 

118 ---------- 

119 config_file : `str` 

120 Name of the configuration file. 

121 

122 Returns 

123 ------- 

124 config : `lsst.ctrl.bps.BpsConfig` 

125 Updated configuration. 

126 qgraph : `lsst.pipe.base.graph.QuantumGraph` 

127 A graph representing quanta. 

128 """ 

129 config = _init_submission_driver(config_file, **kwargs) 

130 submit_path = config[".bps_defined.submitPath"] 

131 

132 _LOG.info("Starting acquire stage (generating and/or reading quantum graph)") 

133 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Acquire stage completed"): 

134 qgraph_file, qgraph, execution_butler_dir = acquire_quantum_graph(config, out_prefix=submit_path) 

135 

136 config[".bps_defined.executionButlerDir"] = execution_butler_dir 

137 config[".bps_defined.runQgraphFile"] = qgraph_file 

138 return config, qgraph 

139 

140 

141def cluster_qgraph_driver(config_file, **kwargs): 

142 """Group quanta into clusters. 

143 

144 Parameters 

145 ---------- 

146 config_file : `str` 

147 Name of the configuration file. 

148 

149 Returns 

150 ------- 

151 config : `lsst.ctrl.bps.BpsConfig` 

152 Updated configuration. 

153 clustered_qgraph : `lsst.ctrl.bps.ClusteredQuantumGraph` 

154 A graph representing clustered quanta. 

155 """ 

156 config, qgraph = acquire_qgraph_driver(config_file, **kwargs) 

157 

158 _LOG.info("Starting cluster stage (grouping quanta into jobs)") 

159 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Cluster stage completed"): 

160 clustered_qgraph = cluster_quanta(config, qgraph, config["uniqProcName"]) 

161 

162 submit_path = config[".bps_defined.submitPath"] 

163 _, save_clustered_qgraph = config.search("saveClusteredQgraph", opt={"default": False}) 

164 if save_clustered_qgraph: 

165 clustered_qgraph.save(os.path.join(submit_path, "bps_clustered_qgraph.pickle")) 

166 _, save_dot = config.search("saveDot", opt={"default": False}) 

167 if save_dot: 

168 clustered_qgraph.draw(os.path.join(submit_path, "bps_clustered_qgraph.dot")) 

169 return config, clustered_qgraph 

170 

171 

172def transform_driver(config_file, **kwargs): 

173 """Create a workflow for a specific workflow management system. 

174 

175 Parameters 

176 ---------- 

177 config_file : `str` 

178 Name of the configuration file. 

179 

180 Returns 

181 ------- 

182 generic_workflow_config : `lsst.ctrl.bps.BpsConfig` 

183 Configuration to use when creating the workflow. 

184 generic_workflow : `lsst.ctrl.bps.BaseWmsWorkflow` 

185 Representation of the abstract/scientific workflow specific to a given 

186 workflow management system. 

187 """ 

188 config, clustered_qgraph = cluster_qgraph_driver(config_file, **kwargs) 

189 submit_path = config[".bps_defined.submitPath"] 

190 

191 _LOG.info("Starting transform stage (creating generic workflow)") 

192 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Transform stage completed"): 

193 generic_workflow, generic_workflow_config = transform(config, clustered_qgraph, submit_path) 

194 _LOG.info("Generic workflow name '%s'", generic_workflow.name) 

195 

196 _, save_workflow = config.search("saveGenericWorkflow", opt={"default": False}) 

197 if save_workflow: 

198 with open(os.path.join(submit_path, "bps_generic_workflow.pickle"), "wb") as outfh: 

199 generic_workflow.save(outfh, "pickle") 

200 _, save_dot = config.search("saveDot", opt={"default": False}) 

201 if save_dot: 

202 with open(os.path.join(submit_path, "bps_generic_workflow.dot"), "w") as outfh: 

203 generic_workflow.draw(outfh, "dot") 

204 return generic_workflow_config, generic_workflow 

205 

206 

207def prepare_driver(config_file, **kwargs): 

208 """Create a representation of the generic workflow. 

209 

210 Parameters 

211 ---------- 

212 config_file : `str` 

213 Name of the configuration file. 

214 

215 Returns 

216 ------- 

217 wms_config : `lsst.ctrl.bps.BpsConfig` 

218 Configuration to use when creating the workflow. 

219 workflow : `lsst.ctrl.bps.BaseWmsWorkflow` 

220 Representation of the abstract/scientific workflow specific to a given 

221 workflow management system. 

222 """ 

223 generic_workflow_config, generic_workflow = transform_driver(config_file, **kwargs) 

224 submit_path = generic_workflow_config[".bps_defined.submitPath"] 

225 

226 _LOG.info("Starting prepare stage (creating specific implementation of workflow)") 

227 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Prepare stage completed"): 

228 wms_workflow = prepare(generic_workflow_config, generic_workflow, submit_path) 

229 

230 wms_workflow_config = generic_workflow_config 

231 print(f"Submit dir: {wms_workflow.submit_path}") 

232 return wms_workflow_config, wms_workflow 

233 

234 

235def submit_driver(config_file, **kwargs): 

236 """Submit workflow for execution. 

237 

238 Parameters 

239 ---------- 

240 config_file : `str` 

241 Name of the configuration file. 

242 """ 

243 _LOG.info("Starting submission process") 

244 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Completed entire submission process"): 

245 wms_workflow_config, wms_workflow = prepare_driver(config_file, **kwargs) 

246 

247 _LOG.info("Starting submit stage") 

248 with time_this(log=_LOG, level=logging.INFO, prefix=None, msg="Completed submit stage"): 

249 submit(wms_workflow_config, wms_workflow) 

250 _LOG.info("Run '%s' submitted for execution with id '%s'", wms_workflow.name, wms_workflow.run_id) 

251 

252 print(f"Run Id: {wms_workflow.run_id}") 

253 

254 

255def report_driver(wms_service, run_id, user, hist_days, pass_thru): 

256 """Print out summary of jobs submitted for execution. 

257 

258 Parameters 

259 ---------- 

260 wms_service : `str` 

261 Name of the class. 

262 run_id : `str` 

263 A run id the report will be restricted to. 

264 user : `str` 

265 A user name the report will be restricted to. 

266 hist_days : int 

267 Number of days 

268 pass_thru : `str` 

269 A string to pass directly to the WMS service class. 

270 """ 

271 report(wms_service, run_id, user, hist_days, pass_thru) 

272 

273 

274def cancel_driver(wms_service, run_id, user, require_bps, pass_thru): 

275 """Cancel submitted workflows. 

276 

277 Parameters 

278 ---------- 

279 wms_service : `str` 

280 Name of the Workload Management System service class. 

281 run_id : `str` 

282 ID or path of job that should be canceled. 

283 user : `str` 

284 User whose submitted jobs should be canceled. 

285 require_bps : `bool` 

286 Whether to require given run_id/user to be a bps submitted job. 

287 pass_thru : `str` 

288 Information to pass through to WMS. 

289 """ 

290 cancel(wms_service, run_id, user, require_bps, pass_thru)