X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/f5f5de0ae41e12738a380c422417d5a5e5af7f09..32d79def3a400de6fbe4cd02a8c71643238d867d:/sdk/cwl/arvados_cwl/runner.py diff --git a/sdk/cwl/arvados_cwl/runner.py b/sdk/cwl/arvados_cwl/runner.py index 69e4f5bd7b..8db3e61cf1 100644 --- a/sdk/cwl/arvados_cwl/runner.py +++ b/sdk/cwl/arvados_cwl/runner.py @@ -1,20 +1,24 @@ +# Copyright (C) The Arvados Authors. All rights reserved. +# +# SPDX-License-Identifier: Apache-2.0 + import os import urlparse from functools import partial import logging import json import subprocess +from collections import namedtuple from StringIO import StringIO -from schema_salad.sourceline import SourceLine +from schema_salad.sourceline import SourceLine, cmap -import cwltool.draft2tool -from cwltool.draft2tool import CommandLineTool +from cwltool.command_line_tool import CommandLineTool import cwltool.workflow from cwltool.process import get_feature, scandeps, UnsupportedRequirement, normalizeFilesDirs, shortname from cwltool.load_tool import fetch_document -from cwltool.pathmapper import adjustFileObjs, adjustDirObjs +from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, visit_class from cwltool.utils import aslist from cwltool.builder import substitute from cwltool.pack import pack @@ -42,8 +46,45 @@ def trim_anonymous_location(obj): if obj.get("location", "").startswith("_:"): del obj["location"] + +def remove_redundant_fields(obj): + for field in ("path", "nameext", "nameroot", "dirname"): + if field in obj: + del obj[field] + + +def find_defaults(d, op): + if isinstance(d, list): + for i in d: + find_defaults(i, op) + elif isinstance(d, dict): + if "default" in d: + op(d) + else: + for i in d.itervalues(): + find_defaults(i, op) + + +def discover_secondary_files(inputs, job_order, discovered=None): + for t in inputs: + def setSecondary(fileobj): + if isinstance(fileobj, dict) and fileobj.get("class") == "File": + if "secondaryFiles" not in fileobj: + fileobj["secondaryFiles"] = cmap([{"location": substitute(fileobj["location"], sf), "class": "File"} for sf in t["secondaryFiles"]]) + if discovered is not None: + discovered[fileobj["location"]] = fileobj["secondaryFiles"] + + if isinstance(fileobj, list): + for e in fileobj: + setSecondary(e) + + if shortname(t["id"]) in job_order and t.get("secondaryFiles"): + setSecondary(job_order[shortname(t["id"])]) + + def upload_dependencies(arvrunner, name, document_loader, - workflowobj, uri, loadref_run, include_primary=True): + workflowobj, uri, loadref_run, + include_primary=True, discovered_secondaryfiles=None): """Upload the dependencies of the workflowobj document to Keep. Returns a pathmapper object mapping local paths to keep references. Also @@ -97,16 +138,50 @@ def upload_dependencies(arvrunner, name, document_loader, for s in workflowobj["$schemas"]: sc.append({"class": "File", "location": s}) + def visit_default(obj): + remove = [False] + def ensure_default_location(f): + if "location" not in f and "path" in f: + f["location"] = f["path"] + del f["path"] + if "location" in f and not arvrunner.fs_access.exists(f["location"]): + # Doesn't exist, remove from list of dependencies to upload + sc[:] = [x for x in sc if x["location"] != f["location"]] + # Delete "default" from workflowobj + remove[0] = True + visit_class(obj["default"], ("File", "Directory"), ensure_default_location) + if remove[0]: + del obj["default"] + + find_defaults(workflowobj, visit_default) + + discovered = {} + def discover_default_secondary_files(obj): + discover_secondary_files(obj["inputs"], + {shortname(t["id"]): t["default"] for t in obj["inputs"] if "default" in t}, + discovered) + + visit_class(workflowobj, ("CommandLineTool", "Workflow"), discover_default_secondary_files) + + for d in discovered: + sc.extend(discovered[d]) + mapper = ArvPathMapper(arvrunner, sc, "", "keep:%s", "keep:%s/%s", - name=name) + name=name, + single_collection=True) def setloc(p): if "location" in p and (not p["location"].startswith("_:")) and (not p["location"].startswith("keep:")): p["location"] = mapper.mapper(p["location"]).resolved - adjustFileObjs(workflowobj, setloc) - adjustDirObjs(workflowobj, setloc) + + visit_class(workflowobj, ("File", "Directory"), setloc) + visit_class(discovered, ("File", "Directory"), setloc) + + if discovered_secondaryfiles is not None: + for d in discovered: + discovered_secondaryfiles[mapper.mapper(d).resolved] = discovered[d] if "$schemas" in workflowobj: sch = [] @@ -123,22 +198,47 @@ def upload_docker(arvrunner, tool): if isinstance(tool, CommandLineTool): (docker_req, docker_is_req) = get_feature(tool, "DockerRequirement") if docker_req: - if docker_req.get("dockerOutputDirectory"): + if docker_req.get("dockerOutputDirectory") and arvrunner.work_api != "containers": # TODO: can be supported by containers API, but not jobs API. raise SourceLine(docker_req, "dockerOutputDirectory", UnsupportedRequirement).makeError( "Option 'dockerOutputDirectory' of DockerRequirement not supported.") arv_docker_get_image(arvrunner.api, docker_req, True, arvrunner.project_uuid) + else: + arv_docker_get_image(arvrunner.api, {"dockerPull": "arvados/jobs"}, True, arvrunner.project_uuid) elif isinstance(tool, cwltool.workflow.Workflow): for s in tool.steps: upload_docker(arvrunner, s.embedded_tool) -def packed_workflow(arvrunner, tool): + +def packed_workflow(arvrunner, tool, merged_map): """Create a packed workflow. A "packed" workflow is one where all the components have been combined into a single document.""" - return pack(tool.doc_loader, tool.doc_loader.fetch(tool.tool["id"]), - tool.tool["id"], tool.metadata) + rewrites = {} + packed = pack(tool.doc_loader, tool.doc_loader.fetch(tool.tool["id"]), + tool.tool["id"], tool.metadata, rewrite_out=rewrites) + + rewrite_to_orig = {} + for k,v in rewrites.items(): + rewrite_to_orig[v] = k + + def visit(v, cur_id): + if isinstance(v, dict): + if v.get("class") in ("CommandLineTool", "Workflow"): + cur_id = rewrite_to_orig.get(v["id"], v["id"]) + if "location" in v and not v["location"].startswith("keep:"): + v["location"] = merged_map[cur_id].resolved[v["location"]] + if "location" in v and v["location"] in merged_map[cur_id].secondaryFiles: + v["secondaryFiles"] = merged_map[cur_id].secondaryFiles[v["location"]] + for l in v: + visit(v[l], cur_id) + if isinstance(v, list): + for l in v: + visit(l, cur_id) + visit(packed, None) + return packed + def tag_git_version(packed): if tool.tool["id"].startswith("file://"): @@ -156,18 +256,7 @@ def upload_job_order(arvrunner, name, tool, job_order): object with 'location' updated to the proper keep references. """ - for t in tool.tool["inputs"]: - def setSecondary(fileobj): - if isinstance(fileobj, dict) and fileobj.get("class") == "File": - if "secondaryFiles" not in fileobj: - fileobj["secondaryFiles"] = [{"location": substitute(fileobj["location"], sf), "class": "File"} for sf in t["secondaryFiles"]] - - if isinstance(fileobj, list): - for e in fileobj: - setSecondary(e) - - if shortname(t["id"]) in job_order and t.get("secondaryFiles"): - setSecondary(job_order[shortname(t["id"])]) + discover_secondary_files(tool.tool["inputs"], job_order) jobmapper = upload_dependencies(arvrunner, name, @@ -186,6 +275,8 @@ def upload_job_order(arvrunner, name, tool, job_order): return job_order +FileUpdates = namedtuple("FileUpdates", ["resolved", "secondaryFiles"]) + def upload_workflow_deps(arvrunner, tool): # Ensure that Docker images needed by this workflow are available @@ -193,19 +284,29 @@ def upload_workflow_deps(arvrunner, tool): document_loader = tool.doc_loader + merged_map = {} + def upload_tool_deps(deptool): if "id" in deptool: - upload_dependencies(arvrunner, - "%s dependencies" % (shortname(deptool["id"])), - document_loader, - deptool, - deptool["id"], - False, - include_primary=False) + discovered_secondaryfiles = {} + pm = upload_dependencies(arvrunner, + "%s dependencies" % (shortname(deptool["id"])), + document_loader, + deptool, + deptool["id"], + False, + include_primary=False, + discovered_secondaryfiles=discovered_secondaryfiles) document_loader.idx[deptool["id"]] = deptool + toolmap = {} + for k,v in pm.items(): + toolmap[k] = v.resolved + merged_map[deptool["id"]] = FileUpdates(toolmap, discovered_secondaryfiles) tool.visit(upload_tool_deps) + return merged_map + def arvados_jobs_image(arvrunner, img): """Determine if the right arvados/jobs image version is available. If not, try to pull and upload it.""" @@ -246,11 +347,19 @@ class Runner(object): def __init__(self, runner, tool, job_order, enable_reuse, output_name, output_tags, submit_runner_ram=0, - name=None, on_error=None, submit_runner_image=None): + name=None, on_error=None, submit_runner_image=None, + intermediate_output_ttl=0, merged_map=None, priority=None, + secret_store=None): self.arvrunner = runner self.tool = tool self.job_order = job_order self.running = False + if enable_reuse: + # If reuse is permitted by command line arguments but + # disabled by the workflow itself, disable it. + reuse_req, _ = get_feature(self.tool, "http://arvados.org/cwl#ReuseRequirement") + if reuse_req: + enable_reuse = reuse_req["enableReuse"] self.enable_reuse = enable_reuse self.uuid = None self.final_output = None @@ -259,6 +368,9 @@ class Runner(object): self.name = name self.on_error = on_error self.jobs_image = submit_runner_image or "arvados/jobs:"+__version__ + self.intermediate_output_ttl = intermediate_output_ttl + self.priority = priority + self.secret_store = secret_store if submit_runner_ram: self.submit_runner_ram = submit_runner_ram @@ -268,6 +380,8 @@ class Runner(object): if self.submit_runner_ram <= 0: raise Exception("Value of --submit-runner-ram must be greater than zero") + self.merged_map = merged_map or {} + def update_pipeline_component(self, record): pass @@ -317,6 +431,3 @@ class Runner(object): self.arvrunner.output_callback({}, "permanentFail") else: self.arvrunner.output_callback(outputs, processStatus) - finally: - if record["uuid"] in self.arvrunner.processes: - del self.arvrunner.processes[record["uuid"]]