X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/2de2c96925cc3439305f16dced7f89bd9124853d..HEAD:/sdk/cwl/arvados_cwl/arvworkflow.py diff --git a/sdk/cwl/arvados_cwl/arvworkflow.py b/sdk/cwl/arvados_cwl/arvworkflow.py index 895676565d..4751e48c04 100644 --- a/sdk/cwl/arvados_cwl/arvworkflow.py +++ b/sdk/cwl/arvados_cwl/arvworkflow.py @@ -2,18 +2,15 @@ # # SPDX-License-Identifier: Apache-2.0 -from past.builtins import basestring -from future.utils import viewitems - import os import json import copy import logging import urllib -from io import StringIO import sys import re +from io import StringIO from typing import (MutableSequence, MutableMapping) from ruamel.yaml import YAML @@ -29,7 +26,7 @@ from cwltool.load_tool import fetch_document, resolve_and_validate_document from cwltool.process import shortname, uniquename from cwltool.workflow import Workflow, WorkflowException, WorkflowStep from cwltool.utils import adjustFileObjs, adjustDirObjs, visit_class, normalizeFilesDirs -from cwltool.context import LoadingContext +from cwltool.context import LoadingContext, getdefault from schema_salad.ref_resolver import file_uri, uri_file_path @@ -38,9 +35,12 @@ import ruamel.yaml as yaml from .runner import (upload_dependencies, packed_workflow, upload_workflow_collection, trim_anonymous_location, remove_redundant_fields, discover_secondary_files, make_builder, arvados_jobs_image, FileUpdates) +from .arvcontainer import RunnerContainer from .pathmapper import ArvPathMapper, trim_listing from .arvtool import ArvadosCommandTool, set_cluster_target from ._version import __version__ +from .util import common_prefix +from .arvdocker import arv_docker_get_image from .perf import Perf @@ -50,6 +50,21 @@ metrics = logging.getLogger('arvados.cwl-runner.metrics') max_res_pars = ("coresMin", "coresMax", "ramMin", "ramMax", "tmpdirMin", "tmpdirMax") sum_res_pars = ("outdirMin", "outdirMax") +_basetype_re = re.compile(r'''(?: +Directory +|File +|array +|boolean +|double +|enum +|float +|int +|long +|null +|record +|string +)(?:\[\])?\??''', re.VERBOSE) + def make_wrapper_workflow(arvRunner, main, packed, project_uuid, name, git_info, tool): col = arvados.collection.Collection(api_client=arvRunner.api, keep_client=arvRunner.keep_client) @@ -130,7 +145,7 @@ def make_wrapper_workflow(arvRunner, main, packed, project_uuid, name, git_info, def rel_ref(s, baseuri, urlexpander, merged_map, jobmapper): - if s.startswith("keep:"): + if s.startswith("keep:") or s.startswith("arvwf:"): return s uri = urlexpander(s, baseuri) @@ -160,21 +175,16 @@ def rel_ref(s, baseuri, urlexpander, merged_map, jobmapper): return os.path.join(r, p3) def is_basetype(tp): - basetypes = ("null", "boolean", "int", "long", "float", "double", "string", "File", "Directory", "record", "array", "enum") - for b in basetypes: - if re.match(b+"(\[\])?\??", tp): - return True - return False + return _basetype_re.match(tp) is not None - -def update_refs(d, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix): +def update_refs(api, d, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix): if isinstance(d, MutableSequence): for i, s in enumerate(d): if prefix and isinstance(s, str): if s.startswith(prefix): d[i] = replacePrefix+s[len(prefix):] else: - update_refs(s, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) + update_refs(api, s, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) elif isinstance(d, MutableMapping): for field in ("id", "name"): if isinstance(d.get(field), str) and d[field].startswith("_:"): @@ -187,8 +197,8 @@ def update_refs(d, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, baseuri = urlexpander(d["name"], baseuri, scoped_id=True) if d.get("class") == "DockerRequirement": - dockerImageId = d.get("dockerImageId") or d.get("dockerPull") - d["http://arvados.org/cwl#dockerCollectionPDH"] = runtimeContext.cached_docker_lookups.get(dockerImageId) + d["http://arvados.org/cwl#dockerCollectionPDH"] = arv_docker_get_image(api, d, False, + runtimeContext) for field in d: if field in ("location", "run", "name") and isinstance(d[field], str): @@ -211,15 +221,21 @@ def update_refs(d, baseuri, urlexpander, merged_map, jobmapper, runtimeContext, if isinstance(d["inputs"][inp], str) and not is_basetype(d["inputs"][inp]): d["inputs"][inp] = rel_ref(d["inputs"][inp], baseuri, urlexpander, merged_map, jobmapper) if isinstance(d["inputs"][inp], MutableMapping): - update_refs(d["inputs"][inp], baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) + update_refs(api, d["inputs"][inp], baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) continue + if field in ("requirements", "hints") and isinstance(d[field], MutableMapping): + dr = d[field].get("DockerRequirement") + if dr: + dr["http://arvados.org/cwl#dockerCollectionPDH"] = arv_docker_get_image(api, dr, False, + runtimeContext) + if field == "$schemas": for n, s in enumerate(d["$schemas"]): d["$schemas"][n] = rel_ref(d["$schemas"][n], baseuri, urlexpander, merged_map, jobmapper) continue - update_refs(d[field], baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) + update_refs(api, d[field], baseuri, urlexpander, merged_map, jobmapper, runtimeContext, prefix, replacePrefix) def fix_schemadef(req, baseuri, urlexpander, merged_map, jobmapper, pdh): @@ -235,6 +251,7 @@ def fix_schemadef(req, baseuri, urlexpander, merged_map, jobmapper, pdh): merged_map[mm].resolved[r] = rename return req + def drop_ids(d): if isinstance(d, MutableSequence): for i, s in enumerate(d): @@ -280,22 +297,8 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, # Find the longest common prefix among all the file names. We'll # use this to recreate the directory structure in a keep # collection with correct relative references. - n = 7 - allmatch = True - if firstfile: - while allmatch: - n += 1 - for f in all_files: - if len(f)-1 < n: - n -= 1 - allmatch = False - break - if f[n] != firstfile[n]: - allmatch = False - break - - while firstfile[n] != "/": - n -= 1 + prefix = common_prefix(firstfile, all_files) if firstfile else "" + col = arvados.collection.Collection(api_client=arvRunner.api) @@ -314,7 +317,7 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, text = tool.doc_loader.fetch_text(w) if isinstance(text, bytes): - textIO = StringIO(text.decode('utf-8')) + textIO = StringIO(str(text, 'utf-8')) else: textIO = StringIO(text) @@ -329,25 +332,25 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, # 2. find $import, $include, $schema, run, location # 3. update field value - update_refs(result, w, tool.doc_loader.expand_url, merged_map, jobmapper, runtimeContext, "", "") + update_refs(arvRunner.api, result, w, tool.doc_loader.expand_url, merged_map, jobmapper, runtimeContext, "", "") # Write the updated file to the collection. - with col.open(w[n+1:], "wt") as f: + with col.open(w[len(prefix):], "wt") as f: if export_as_json: json.dump(result, f, indent=4, separators=(',',': ')) else: yamlloader.dump(result, stream=f) # Also store a verbatim copy of the original files - with col.open(os.path.join("original", w[n+1:]), "wt") as f: + with col.open(os.path.join("original", w[len(prefix):]), "wt") as f: f.write(text) # Upload files referenced by $include directives, these are used # unchanged and don't need to be updated. for w in include_files: - with col.open(w[n+1:], "wb") as f1: - with col.open(os.path.join("original", w[n+1:]), "wb") as f3: + with col.open(w[len(prefix):], "wb") as f1: + with col.open(os.path.join("original", w[len(prefix):]), "wb") as f3: with open(uri_file_path(w), "rb") as f2: dat = f2.read(65536) while dat: @@ -361,7 +364,7 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, if git_info and git_info.get("http://arvados.org/cwl#gitDescribe"): toolname = "%s (%s)" % (toolname, git_info.get("http://arvados.org/cwl#gitDescribe")) - toolfile = tool.tool["id"][n+1:] + toolfile = tool.tool["id"][len(prefix):] properties = { "type": "workflow", @@ -414,9 +417,10 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, wf_runner_resources = {"class": "http://arvados.org/cwl#WorkflowRunnerResources"} hints.append(wf_runner_resources) - wf_runner_resources["acrContainerImage"] = arvados_jobs_image(arvRunner, - submit_runner_image or "arvados/jobs:"+__version__, - runtimeContext) + if "acrContainerImage" not in wf_runner_resources: + wf_runner_resources["acrContainerImage"] = arvados_jobs_image(arvRunner, + submit_runner_image or "arvados/jobs:"+__version__, + runtimeContext) if submit_runner_ram: wf_runner_resources["ramMin"] = submit_runner_ram @@ -486,7 +490,7 @@ def upload_workflow(arvRunner, tool, job_order, project_uuid, if r["class"] == "SchemaDefRequirement": wrapper["requirements"][i] = fix_schemadef(r, main["id"], tool.doc_loader.expand_url, merged_map, jobmapper, col.portable_data_hash()) - update_refs(wrapper, main["id"], tool.doc_loader.expand_url, merged_map, jobmapper, runtimeContext, main["id"]+"#", "#main/") + update_refs(arvRunner.api, wrapper, main["id"], tool.doc_loader.expand_url, merged_map, jobmapper, runtimeContext, main["id"]+"#", "#main/") doc = {"cwlVersion": "v1.2", "$graph": [wrapper]} @@ -581,7 +585,7 @@ class ArvadosWorkflowStep(WorkflowStep): runtimeContext = runtimeContext.copy() runtimeContext.toplevel = True # Preserve behavior for #13365 - builder = make_builder({shortname(k): v for k,v in viewitems(joborder)}, self.hints, self.requirements, + builder = make_builder({shortname(k): v for k, v in joborder.items()}, self.hints, self.requirements, runtimeContext, self.metadata) runtimeContext = set_cluster_target(self.tool, self.arvrunner, builder, runtimeContext) return super(ArvadosWorkflowStep, self).job(joborder, output_callback, runtimeContext) @@ -596,21 +600,22 @@ class ArvadosWorkflow(Workflow): self.dynamic_resource_req = [] self.static_resource_req = [] self.wf_reffiles = [] - self.loadingContext = loadingContext - super(ArvadosWorkflow, self).__init__(toolpath_object, loadingContext) - self.cluster_target_req, _ = self.get_requirement("http://arvados.org/cwl#ClusterTarget") + self.loadingContext = loadingContext.copy() - def job(self, joborder, output_callback, runtimeContext): + self.requirements = copy.deepcopy(getdefault(loadingContext.requirements, [])) + tool_requirements = toolpath_object.get("requirements", []) + self.hints = copy.deepcopy(getdefault(loadingContext.hints, [])) + tool_hints = toolpath_object.get("hints", []) - builder = make_builder(joborder, self.hints, self.requirements, runtimeContext, self.metadata) - runtimeContext = set_cluster_target(self.tool, self.arvrunner, builder, runtimeContext) + workflow_runner_req, _ = self.get_requirement("http://arvados.org/cwl#WorkflowRunnerResources") + if workflow_runner_req and workflow_runner_req.get("acrContainerImage"): + self.loadingContext.default_docker_image = workflow_runner_req.get("acrContainerImage") - req, _ = self.get_requirement("http://arvados.org/cwl#RunInSingleContainer") - if not req: - return super(ArvadosWorkflow, self).job(joborder, output_callback, runtimeContext) + super(ArvadosWorkflow, self).__init__(toolpath_object, self.loadingContext) + self.cluster_target_req, _ = self.get_requirement("http://arvados.org/cwl#ClusterTarget") - # RunInSingleContainer is true + def runInSingleContainer(self, joborder, output_callback, runtimeContext, builder): with SourceLine(self.tool, None, WorkflowException, logger.isEnabledFor(logging.DEBUG)): if "id" not in self.tool: raise WorkflowException("%s object must have 'id'" % (self.tool["class"])) @@ -647,7 +652,7 @@ class ArvadosWorkflow(Workflow): dyn = False for k in max_res_pars + sum_res_pars: if k in req: - if isinstance(req[k], basestring): + if isinstance(req[k], str): if item["id"] == "#main": # only the top-level requirements/hints may contain expressions self.dynamic_resource_req.append(req) @@ -773,6 +778,51 @@ class ArvadosWorkflow(Workflow): }) return ArvadosCommandTool(self.arvrunner, wf_runner, self.loadingContext).job(joborder_resolved, output_callback, runtimeContext) + + def separateRunner(self, joborder, output_callback, runtimeContext, req, builder): + + name = runtimeContext.name + + rpn = req.get("runnerProcessName") + if rpn: + name = builder.do_eval(rpn) + + return RunnerContainer(self.arvrunner, + self, + self.loadingContext, + runtimeContext.enable_reuse, + None, + None, + submit_runner_ram=runtimeContext.submit_runner_ram, + name=name, + on_error=runtimeContext.on_error, + submit_runner_image=runtimeContext.submit_runner_image, + intermediate_output_ttl=runtimeContext.intermediate_output_ttl, + merged_map=None, + priority=runtimeContext.priority, + secret_store=self.arvrunner.secret_store, + collection_cache_size=runtimeContext.collection_cache_size, + collection_cache_is_default=self.arvrunner.should_estimate_cache_size, + git_info=runtimeContext.git_info, + reuse_runner=True).job(joborder, output_callback, runtimeContext) + + + def job(self, joborder, output_callback, runtimeContext): + + builder = make_builder(joborder, self.hints, self.requirements, runtimeContext, self.metadata) + runtimeContext = set_cluster_target(self.tool, self.arvrunner, builder, runtimeContext) + + req, _ = self.get_requirement("http://arvados.org/cwl#RunInSingleContainer") + if req: + return self.runInSingleContainer(joborder, output_callback, runtimeContext, builder) + + req, _ = self.get_requirement("http://arvados.org/cwl#SeparateRunner") + if req: + return self.separateRunner(joborder, output_callback, runtimeContext, req, builder) + + return super(ArvadosWorkflow, self).job(joborder, output_callback, runtimeContext) + + def make_workflow_step(self, toolpath_object, # type: Dict[Text, Any] pos, # type: int