X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/0939c5719af9601ae6b95f6d230cbc4cace122cc..a3b40d4546cb18744d6052ffb47f5f81773ac261:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index d9466a83a0..b0dff491f5 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -2,17 +2,21 @@ # # SPDX-License-Identifier: Apache-2.0 +from future import standard_library +standard_library.install_aliases() +from builtins import str + import logging import json import os -import urllib +import urllib.request, urllib.parse, urllib.error import time import datetime import ciso8601 import uuid import math -from arvados_cwl.util import get_current_container, get_intermediate_collection_info +import arvados_cwl.util import ruamel.yaml as yaml from cwltool.errors import WorkflowException @@ -36,7 +40,7 @@ metrics = logging.getLogger('arvados.cwl-runner.metrics') class ArvadosContainer(JobBase): """Submit and manage a Crunch container request for executing a CWL CommandLineTool.""" - def __init__(self, runner, + def __init__(self, runner, job_runtime, builder, # type: Builder joborder, # type: Dict[Text, Union[Dict[Text, Any], List, Text]] make_path_mapper, # type: Callable[..., PathMapper] @@ -46,6 +50,7 @@ class ArvadosContainer(JobBase): ): super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name) self.arvrunner = runner + self.job_runtime = job_runtime self.running = False self.uuid = None @@ -60,6 +65,8 @@ class ArvadosContainer(JobBase): # ArvadosContainer object by CommandLineTool.job() before # run() is called. + runtimeContext = self.job_runtime + container_request = { "command": self.command_line, "name": self.name, @@ -71,8 +78,8 @@ class ArvadosContainer(JobBase): } runtime_constraints = {} - if self.arvrunner.project_uuid: - container_request["owner_uuid"] = self.arvrunner.project_uuid + if runtimeContext.project_uuid: + container_request["owner_uuid"] = runtimeContext.project_uuid if self.arvrunner.secret_store.has_secret(self.command_line): raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets") @@ -116,6 +123,8 @@ class ArvadosContainer(JobBase): "kind": "collection", "portable_data_hash": pdh } + if pdh in self.pathmapper.pdh_to_uuid: + mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh] if len(sp) == 2: if tp == "Directory": path = sp[1] @@ -130,10 +139,10 @@ class ArvadosContainer(JobBase): vwd = arvados.collection.Collection(api_client=self.arvrunner.api, keep_client=self.arvrunner.keep_client, num_retries=self.arvrunner.num_retries) - generatemapper = NoFollowPathMapper([self.generatefiles], "", "", + generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "", separateDirs=False) - sorteditems = sorted(generatemapper.items(), None, key=lambda n: n[1].target) + sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target) logger.debug("generatemapper is %s", sorteditems) @@ -155,7 +164,7 @@ class ArvadosContainer(JobBase): } else: with vwd.open(p.target, "w") as n: - n.write(p.resolved.encode("utf-8")) + n.write(p.resolved) def keepemptydirs(p): if isinstance(p, arvados.collection.RichCollectionBase): @@ -168,10 +177,10 @@ class ArvadosContainer(JobBase): keepemptydirs(vwd) if not runtimeContext.current_container: - runtimeContext.current_container = get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger) - info = get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl) + runtimeContext.current_container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger) + info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl) vwd.save_new(name=info["name"], - owner_uuid=self.arvrunner.project_uuid, + owner_uuid=runtimeContext.project_uuid, ensure_unique_name=True, trash_at=info["trash_at"], properties=info["properties"]) @@ -214,7 +223,7 @@ class ArvadosContainer(JobBase): container_request["container_image"] = arv_docker_get_image(self.arvrunner.api, docker_req, runtimeContext.pull_image, - self.arvrunner.project_uuid) + runtimeContext.project_uuid) api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement") if api_req: @@ -251,13 +260,8 @@ class ArvadosContainer(JobBase): scheduling_parameters["max_run_time"] = self.timelimit extra_submit_params = {} - cluster_target_req, _ = self.get_requirement("http://arvados.org/cwl#ClusterTarget") - if cluster_target_req: - cluster_id = cluster_target_req.get("clusterID") - if cluster_id: - extra_submit_params["cluster_id"] = cluster_id - if cluster_target_req.get("ownerUUID"): - container_request["owner_uuid"] = cluster_target_req.get("ownerUUID") + if runtimeContext.submit_runner_cluster: + extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster container_request["output_name"] = "Output for step %s" % (self.name) container_request["output_ttl"] = self.output_ttl @@ -302,8 +306,8 @@ class ArvadosContainer(JobBase): logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"]) else: logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"]) - except Exception as e: - logger.error("%s got error %s" % (self.arvrunner.label(self), str(e))) + except Exception: + logger.exception("%s got an error", self.arvrunner.label(self)) self.output_callback({}, "permanentFail") def done(self, record): @@ -327,8 +331,8 @@ class ArvadosContainer(JobBase): else: processStatus = "permanentFail" - if processStatus == "permanentFail": - logc = arvados.collection.CollectionReader(container["log"], + if processStatus == "permanentFail" and record["log_uuid"]: + logc = arvados.collection.CollectionReader(record["log_uuid"], api_client=self.arvrunner.api, keep_client=self.arvrunner.keep_client, num_retries=self.arvrunner.num_retries) @@ -340,7 +344,7 @@ class ArvadosContainer(JobBase): if record["output_uuid"]: if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl: # Compute the trash time to avoid requesting the collection record. - trash_at = ciso8601.parse_datetime_unaware(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl) + trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl) aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else "" orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else "" oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else "" @@ -351,11 +355,13 @@ class ArvadosContainer(JobBase): if container["output"]: outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep") except WorkflowException as e: + # Only include a stack trace if in debug mode. + # A stack trace may obfuscate more useful output about the workflow. logger.error("%s unable to collect output from %s:\n%s", self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False)) processStatus = "permanentFail" - except Exception as e: - logger.exception("%s while getting output object: %s", self.arvrunner.label(self), e) + except Exception: + logger.exception("%s while getting output object:", self.arvrunner.label(self)) processStatus = "permanentFail" finally: self.output_callback(outputs, processStatus) @@ -409,15 +415,15 @@ class RunnerContainer(Runner): "secret_mounts": secret_mounts, "runtime_constraints": { "vcpus": math.ceil(self.submit_runner_cores), - "ram": math.ceil(1024*1024 * self.submit_runner_ram), + "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)), "API": True }, "use_existing": self.enable_reuse, "properties": {} } - if self.tool.tool.get("id", "").startswith("keep:"): - sp = self.tool.tool["id"].split('/') + if self.embedded_tool.tool.get("id", "").startswith("keep:"): + sp = self.embedded_tool.tool["id"].split('/') workflowcollection = sp[0][5:] workflowname = "/".join(sp[1:]) workflowpath = "/var/lib/cwl/workflow/%s" % workflowname @@ -426,14 +432,14 @@ class RunnerContainer(Runner): "portable_data_hash": "%s" % workflowcollection } else: - packed = packed_workflow(self.arvrunner, self.tool, self.merged_map) + packed = packed_workflow(self.arvrunner, self.embedded_tool, self.merged_map) workflowpath = "/var/lib/cwl/workflow.json#main" container_req["mounts"]["/var/lib/cwl/workflow.json"] = { "kind": "json", "content": packed } - if self.tool.tool.get("id", "").startswith("arvwf:"): - container_req["properties"]["template_uuid"] = self.tool.tool["id"][6:33] + if self.embedded_tool.tool.get("id", "").startswith("arvwf:"): + container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33] # --local means execute the workflow instead of submitting a container request @@ -443,6 +449,7 @@ class RunnerContainer(Runner): # --eval-timeout is the timeout for javascript invocation # --parallel-task-count is the number of threads to use for job submission # --enable/disable-reuse sets desired job reuse + # --collection-cache-size sets aside memory to store collections command = ["arvados-cwl-runner", "--local", "--api=containers", @@ -450,7 +457,8 @@ class RunnerContainer(Runner): "--disable-validate", "--eval-timeout=%s" % self.arvrunner.eval_timeout, "--thread-count=%s" % self.arvrunner.thread_count, - "--enable-reuse" if self.enable_reuse else "--disable-reuse"] + "--enable-reuse" if self.enable_reuse else "--disable-reuse", + "--collection-cache-size=%s" % self.collection_cache_size] if self.output_name: command.append("--output-name=" + self.output_name) @@ -477,6 +485,9 @@ class RunnerContainer(Runner): if self.arvrunner.project_uuid: command.append("--project-uuid="+self.arvrunner.project_uuid) + if self.enable_dev: + command.append("--enable-dev") + command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"]) container_req["command"] = command @@ -495,6 +506,9 @@ class RunnerContainer(Runner): extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster if runtimeContext.submit_request_uuid: + if "cluster_id" in extra_submit_params: + # Doesn't make sense for "update" and actually fails + del extra_submit_params["cluster_id"] response = self.arvrunner.api.container_requests().update( uuid=runtimeContext.submit_request_uuid, body=job_spec, @@ -516,8 +530,9 @@ class RunnerContainer(Runner): container = self.arvrunner.api.containers().get( uuid=record["container_uuid"] ).execute(num_retries=self.arvrunner.num_retries) - except Exception as e: - logger.exception("%s while getting runner container: %s", self.arvrunner.label(self), e) + container["log"] = record["log_uuid"] + except Exception: + logger.exception("%s while getting runner container", self.arvrunner.label(self)) self.arvrunner.output_callback({}, "permanentFail") else: super(RunnerContainer, self).done(container)