2 # Copyright (C) The Arvados Authors. All rights reserved.
4 # SPDX-License-Identifier: Apache-2.0
6 # Implement cwl-runner interface for submitting and running work on Arvados, using
7 # the Crunch containers API.
10 import importlib.metadata
11 import importlib.resources
17 from schema_salad.sourceline import SourceLine
18 import schema_salad.validate as validate
20 import cwltool.workflow
21 import cwltool.process
22 import cwltool.argparser
23 from cwltool.errors import WorkflowException
24 from cwltool.process import shortname, UnsupportedRequirement, use_custom_schema
25 from cwltool.utils import adjustFileObjs, adjustDirObjs, get_listing
29 import arvados.logging
30 from arvados.keep import KeepClient
31 from arvados.errors import ApiError
32 import arvados.commands._util as arv_cmd
34 from .perf import Perf
35 from ._version import __version__
36 from .executor import ArvCwlExecutor
37 from .fsaccess import workflow_uuid_pattern
39 # These aren't used directly in this file but
40 # other code expects to import them from here
41 from .arvcontainer import ArvadosContainer
42 from .arvtool import ArvadosCommandTool
43 from .fsaccess import CollectionFsAccess, CollectionCache, CollectionFetcher
44 from .util import get_current_container
45 from .executor import RuntimeStatusLoggingHandler, DEFAULT_PRIORITY
46 from .arvworkflow import ArvadosWorkflow
48 logger = logging.getLogger('arvados.cwl-runner')
49 metrics = logging.getLogger('arvados.cwl-runner.metrics')
50 logger.setLevel(logging.INFO)
52 arvados.log_handler.setFormatter(logging.Formatter(
53 '%(asctime)s %(name)s %(levelname)s: %(message)s',
57 """Print version string of key packages for provenance and debugging."""
58 return "{} {}, arvados-python-client {}, cwltool {}".format(
60 importlib.metadata.version('arvados-cwl-runner'),
61 importlib.metadata.version('arvados-python-client'),
62 importlib.metadata.version('cwltool'),
65 def arg_parser(): # type: () -> argparse.ArgumentParser
66 parser = argparse.ArgumentParser(
67 description='Arvados executor for Common Workflow Language',
68 parents=[arv_cmd.retry_opt],
71 parser.add_argument("--basedir",
72 help="Base directory used to resolve relative references in the input, default to directory of input object file or current directory (if inputs piped/provided on command line).")
73 parser.add_argument("--outdir", default=os.path.abspath('.'),
74 help="Output directory, default current directory")
76 parser.add_argument("--eval-timeout",
77 help="Time to wait for a Javascript expression to evaluate before giving an error, default 20s.",
81 exgroup = parser.add_mutually_exclusive_group()
82 exgroup.add_argument("--print-dot", action="store_true",
83 help="Print workflow visualization in graphviz format and exit")
84 exgroup.add_argument("--version", action="version", help="Print version and exit", version=versionstring())
85 exgroup.add_argument("--validate", action="store_true", help="Validate CWL document only.")
87 exgroup = parser.add_mutually_exclusive_group()
88 exgroup.add_argument("--verbose", action="store_true", help="Default logging")
89 exgroup.add_argument("--quiet", action="store_true", help="Only print warnings and errors.")
90 exgroup.add_argument("--debug", action="store_true", help="Print even more logging")
92 parser.add_argument("--metrics", action="store_true", help="Print timing metrics")
94 parser.add_argument("--tool-help", action="store_true", help="Print command line help for tool")
96 exgroup = parser.add_mutually_exclusive_group()
97 exgroup.add_argument("--enable-reuse", action="store_true",
98 default=True, dest="enable_reuse",
99 help="Enable container reuse (default)")
100 exgroup.add_argument("--disable-reuse", action="store_false",
101 default=True, dest="enable_reuse",
102 help="Disable container reuse")
104 parser.add_argument("--project-uuid", metavar="UUID", help="Project that will own the workflow containers, if not provided, will go to home project.")
105 parser.add_argument("--output-name", help="Name to use for collection that stores the final output.", default=None)
106 parser.add_argument("--output-tags", help="Tags for the final output collection separated by commas, e.g., '--output-tags tag0,tag1,tag2'.", default=None)
107 parser.add_argument("--ignore-docker-for-reuse", action="store_true",
108 help="Ignore Docker image version when deciding whether to reuse past containers.",
111 exgroup = parser.add_mutually_exclusive_group()
112 exgroup.add_argument("--submit", action="store_true", help="Submit workflow to run on Arvados.",
113 default=True, dest="submit")
114 exgroup.add_argument("--local", action="store_false", help="Run workflow on local host (submits containers to Arvados).",
115 default=True, dest="submit")
116 exgroup.add_argument("--create-template", action="store_true", help="(Deprecated) synonym for --create-workflow.",
117 dest="create_workflow")
118 exgroup.add_argument("--create-workflow", action="store_true", help="Register an Arvados workflow that can be run from Workbench")
119 exgroup.add_argument("--update-workflow", metavar="UUID", help="Update an existing Arvados workflow with the given UUID.")
121 exgroup.add_argument("--print-keep-deps", action="store_true", help="To assist copying, print a list of Keep collections that this workflow depends on.")
123 exgroup = parser.add_mutually_exclusive_group()
124 exgroup.add_argument("--wait", action="store_true", help="After submitting workflow runner, wait for completion.",
125 default=True, dest="wait")
126 exgroup.add_argument("--no-wait", action="store_false", help="Submit workflow runner and exit.",
127 default=True, dest="wait")
129 exgroup = parser.add_mutually_exclusive_group()
130 exgroup.add_argument("--log-timestamps", action="store_true", help="Prefix logging lines with timestamp",
131 default=True, dest="log_timestamps")
132 exgroup.add_argument("--no-log-timestamps", action="store_false", help="No timestamp on logging lines",
133 default=True, dest="log_timestamps")
135 parser.add_argument("--api",
136 default=None, dest="work_api",
137 choices=("containers",),
138 help="Select work submission API. Only supports 'containers'")
140 parser.add_argument("--compute-checksum", action="store_true", default=False,
141 help="Compute checksum of contents while collecting outputs",
142 dest="compute_checksum")
144 parser.add_argument("--submit-runner-ram", type=int,
145 help="RAM (in MiB) required for the workflow runner job (default 1024)",
148 parser.add_argument("--submit-runner-image",
149 help="Docker image for workflow runner job, default arvados/jobs:%s" % __version__,
152 parser.add_argument("--always-submit-runner", action="store_true",
153 help="When invoked with --submit --wait, always submit a runner to manage the workflow, even when only running a single CommandLineTool",
156 parser.add_argument("--match-submitter-images", action="store_true",
157 default=False, dest="match_local_docker",
158 help="Where Arvados has more than one Docker image of the same name, use image from the Docker instance on the submitting node.")
160 exgroup = parser.add_mutually_exclusive_group()
161 exgroup.add_argument("--submit-request-uuid",
163 help="Update and commit to supplied container request instead of creating a new one.",
165 exgroup.add_argument("--submit-runner-cluster",
166 help="Submit workflow runner to a remote cluster",
168 metavar="CLUSTER_ID")
170 parser.add_argument("--collection-cache-size", type=int,
172 help="Collection cache size (in MiB, default 256).")
174 parser.add_argument("--name",
175 help="Name to use for workflow execution instance.",
178 parser.add_argument("--on-error",
179 help="Desired workflow behavior when a step fails. One of 'stop' (do not submit any more steps) or "
180 "'continue' (may submit other steps that are not downstream from the error). Default is 'continue'.",
181 default="continue", choices=("stop", "continue"))
183 parser.add_argument("--enable-dev", action="store_true",
184 help="Enable loading and running development versions "
185 "of the CWL standards.", default=False)
186 parser.add_argument('--storage-classes', default="default",
187 help="Specify comma separated list of storage classes to be used when saving final workflow output to Keep.")
188 parser.add_argument('--intermediate-storage-classes', default="default",
189 help="Specify comma separated list of storage classes to be used when saving intermediate workflow output to Keep.")
191 parser.add_argument("--intermediate-output-ttl", type=int, metavar="N",
192 help="If N > 0, intermediate output collections will be trashed N seconds after creation. Default is 0 (don't trash).",
195 parser.add_argument("--priority", type=int,
196 help="Workflow priority (range 1..1000, higher has precedence over lower)",
197 default=DEFAULT_PRIORITY)
199 parser.add_argument("--disable-validate", dest="do_validate",
200 action="store_false", default=True,
201 help=argparse.SUPPRESS)
203 parser.add_argument("--disable-git", dest="git_info",
204 action="store_false", default=True,
205 help=argparse.SUPPRESS)
207 parser.add_argument("--disable-color", dest="enable_color",
208 action="store_false", default=True,
209 help=argparse.SUPPRESS)
211 parser.add_argument("--disable-js-validation",
212 action="store_true", default=False,
213 help=argparse.SUPPRESS)
215 parser.add_argument("--fast-parser", dest="fast_parser",
216 action="store_true", default=False,
217 help=argparse.SUPPRESS)
219 parser.add_argument("--thread-count", type=int,
220 default=0, help="Number of threads to use for job submit and output collection.")
222 parser.add_argument("--http-timeout", type=int,
223 default=5*60, dest="http_timeout", help="API request timeout in seconds. Default is 300 seconds (5 minutes).")
225 parser.add_argument("--defer-downloads", action="store_true", default=False,
226 help="When submitting a workflow, defer downloading HTTP URLs to workflow launch instead of downloading to Keep before submit.")
228 parser.add_argument("--varying-url-params", type=str, default="",
229 help="A comma separated list of URL query parameters that should be ignored when storing HTTP URLs in Keep.")
231 parser.add_argument("--prefer-cached-downloads", action="store_true", default=False,
232 help="If a HTTP URL is found in Keep, skip upstream URL freshness check (will not notice if the upstream has changed, but also not error if upstream is unavailable).")
234 exgroup = parser.add_mutually_exclusive_group()
235 exgroup.add_argument("--enable-preemptible", dest="enable_preemptible", default=None, action="store_true", help="Use preemptible instances. Control individual steps with arv:UsePreemptible hint.")
236 exgroup.add_argument("--disable-preemptible", dest="enable_preemptible", default=None, action="store_false", help="Don't use preemptible instances.")
238 exgroup = parser.add_mutually_exclusive_group()
239 exgroup.add_argument("--enable-resubmit-non-preemptible", dest="enable_resubmit_non_preemptible",
240 default=None, action="store_true",
241 help="If a workflow step fails due to the instance it is running on being preempted, re-submit the container with the `preemptible` flag disabled. Control individual steps with arv:PreemptionBehavior hint.")
242 exgroup.add_argument("--disable-resubmit-non-preemptible", dest="enable_resubmit_non_preemptible",
243 default=None, action="store_false",
244 help="Don't resumbit when a preemptible instance is reclaimed.")
246 exgroup = parser.add_mutually_exclusive_group()
247 exgroup.add_argument("--copy-deps", dest="copy_deps", default=None, action="store_true", help="Copy dependencies into the destination project.")
248 exgroup.add_argument("--no-copy-deps", dest="copy_deps", default=None, action="store_false", help="Leave dependencies where they are.")
253 help="Skip loading of schemas",
258 exgroup = parser.add_mutually_exclusive_group()
259 exgroup.add_argument("--trash-intermediate", action="store_true",
260 default=False, dest="trash_intermediate",
261 help="Immediately trash intermediate outputs on workflow success.")
262 exgroup.add_argument("--no-trash-intermediate", action="store_false",
263 default=False, dest="trash_intermediate",
264 help="Do not trash intermediate outputs (default).")
266 exgroup = parser.add_mutually_exclusive_group()
267 exgroup.add_argument("--enable-usage-report", dest="enable_usage_report", default=None, action="store_true", help="Create usage_report.html with a summary of each step's resource usage.")
268 exgroup.add_argument("--disable-usage-report", dest="enable_usage_report", default=None, action="store_false", help="Disable usage report.")
270 parser.add_argument("workflow", default=None, help="The workflow to execute")
271 parser.add_argument("job_order", nargs=argparse.REMAINDER, help="The input object to the workflow.")
276 cwltool.command_line_tool.ACCEPTLIST_EN_RELAXED_RE = re.compile(r".*")
277 cwltool.command_line_tool.ACCEPTLIST_RE = cwltool.command_line_tool.ACCEPTLIST_EN_RELAXED_RE
278 supported_versions = ["v1.0", "v1.1", "v1.2"]
279 for s in supported_versions:
280 customschema = importlib.resources.read_text(__name__, f'arv-cwl-schema-{s}.yml', encoding='utf-8')
281 use_custom_schema(s, "http://arvados.org/cwl", customschema)
282 cwltool.process.supportedProcessRequirements.extend([
283 "http://arvados.org/cwl#RunInSingleContainer",
284 "http://arvados.org/cwl#OutputDirType",
285 "http://arvados.org/cwl#RuntimeConstraints",
286 "http://arvados.org/cwl#PartitionRequirement",
287 "http://arvados.org/cwl#APIRequirement",
288 "http://commonwl.org/cwltool#LoadListingRequirement",
289 "http://arvados.org/cwl#IntermediateOutput",
290 "http://arvados.org/cwl#ReuseRequirement",
291 "http://arvados.org/cwl#ClusterTarget",
292 "http://arvados.org/cwl#OutputStorageClass",
293 "http://arvados.org/cwl#ProcessProperties",
294 "http://commonwl.org/cwltool#CUDARequirement",
295 "http://arvados.org/cwl#UsePreemptible",
296 "http://arvados.org/cwl#OutputCollectionProperties",
297 "http://arvados.org/cwl#KeepCacheTypeRequirement",
298 "http://arvados.org/cwl#OutOfMemoryRetry",
299 "http://arvados.org/cwl#PreemptionBehavior",
302 def exit_signal_handler(sigcode, frame):
303 logger.error(str(u"Caught signal {}, exiting.").format(sigcode))
306 def main(args=sys.argv[1:],
311 install_sig_handlers=True):
312 parser = arg_parser()
314 job_order_object = None
315 arvargs = parser.parse_args(args)
317 arvargs.use_container = True
318 arvargs.relax_path_checks = True
319 arvargs.print_supported_versions = False
321 if install_sig_handlers:
322 arv_cmd.install_signal_handlers()
324 if arvargs.update_workflow:
325 if arvargs.update_workflow.find('-7fd4e-') == 5:
326 want_api = 'containers'
329 if want_api and arvargs.work_api and want_api != arvargs.work_api:
330 logger.error(str(u'--update-workflow arg {!r} uses {!r} API, but --api={!r} specified').format(
331 arvargs.update_workflow, want_api, arvargs.work_api))
333 arvargs.work_api = want_api
335 workflow_op = arvargs.create_workflow or arvargs.update_workflow or arvargs.print_keep_deps
337 if workflow_op and not arvargs.job_order:
338 job_order_object = ({}, "")
342 for key, val in cwltool.argparser.get_default_args().items():
343 if not hasattr(arvargs, key):
344 setattr(arvargs, key, val)
347 if api_client is None:
348 api_client = arvados.safeapi.ThreadSafeApiCache(
350 'num_retries': arvargs.retries,
351 'timeout': arvargs.http_timeout,
354 'num_retries': arvargs.retries,
358 keep_client = api_client.keep
359 # Make an API object now so errors are reported early.
360 api_client.users().current().execute()
361 if keep_client is None:
362 block_cache = arvados.keep.KeepBlockCache(disk_cache=True)
363 keep_client = arvados.keep.KeepClient(
364 api_client=api_client,
365 block_cache=block_cache,
366 num_retries=arvargs.retries,
368 executor = ArvCwlExecutor(
371 keep_client=keep_client,
372 num_retries=arvargs.retries,
375 except WorkflowException as e:
376 logger.error(e, exc_info=(sys.exc_info()[1] if arvargs.debug else False))
379 logger.exception("Error creating the Arvados CWL Executor")
382 # Note that unless in debug mode, some stack traces related to user
383 # workflow errors may be suppressed.
385 # Set the logging on most modules INFO (instead of default which is WARNING)
386 logger.setLevel(logging.INFO)
387 logging.getLogger('arvados').setLevel(logging.INFO)
388 logging.getLogger('arvados.keep').setLevel(logging.WARNING)
389 # API retries are filtered to the INFO level and can be noisy, but as long as
390 # they succeed we don't need to see warnings about it.
391 googleapiclient_http_logger = logging.getLogger('googleapiclient.http')
392 googleapiclient_http_logger.addFilter(arvados.logging.GoogleHTTPClientFilter())
393 googleapiclient_http_logger.setLevel(logging.WARNING)
396 logger.setLevel(logging.DEBUG)
397 logging.getLogger('arvados').setLevel(logging.DEBUG)
398 # In debug mode show logs about retries, but we arn't
399 # debugging the google client so we don't need to see
401 googleapiclient_http_logger.setLevel(logging.NOTSET)
402 logging.getLogger('googleapiclient').setLevel(logging.INFO)
405 logger.setLevel(logging.WARN)
406 logging.getLogger('arvados').setLevel(logging.WARN)
407 logging.getLogger('arvados.arv-run').setLevel(logging.WARN)
410 metrics.setLevel(logging.DEBUG)
411 logging.getLogger("cwltool.metrics").setLevel(logging.DEBUG)
413 if arvargs.log_timestamps:
414 arvados.log_handler.setFormatter(logging.Formatter(
415 '%(asctime)s %(name)s %(levelname)s: %(message)s',
416 '%Y-%m-%d %H:%M:%S'))
418 arvados.log_handler.setFormatter(logging.Formatter('%(name)s %(levelname)s: %(message)s'))
420 if stdout is sys.stdout:
421 # cwltool.main has code to work around encoding issues with
422 # sys.stdout and unix pipes (they default to ASCII encoding,
423 # we want utf-8), so when stdout is sys.stdout set it to None
424 # to take advantage of that. Don't override it for all cases
425 # since we still want to be able to capture stdout for the
429 executor.loadingContext.default_docker_image = arvargs.submit_runner_image or "arvados/jobs:"+__version__
431 if arvargs.workflow.startswith("arvwf:") or workflow_uuid_pattern.match(arvargs.workflow) or arvargs.workflow.startswith("keep:"):
432 executor.loadingContext.do_validate = False
433 if arvargs.submit and not workflow_op:
434 executor.fast_submit = True
436 return cwltool.main.main(args=arvargs,
439 executor=executor.arv_executor,
440 versionfunc=versionstring,
441 job_order_object=job_order_object,
442 logger_handler=arvados.log_handler,
443 custom_schema_callback=add_arv_hints,
444 loadingContext=executor.loadingContext,
445 runtimeContext=executor.toplevel_runtimeContext,
446 input_required=not workflow_op)