X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/aea835bc965d42e225c2641b0210c4b521f6dc4e..80459d52161120ae8e33da140984d596271d5195:/sdk/cwl/arvados_cwl/arvjob.py

diff --git a/sdk/cwl/arvados_cwl/arvjob.py b/sdk/cwl/arvados_cwl/arvjob.py
index 87bacd02d3..877f4238a3 100644
--- a/sdk/cwl/arvados_cwl/arvjob.py
+++ b/sdk/cwl/arvados_cwl/arvjob.py
@@ -9,17 +9,18 @@ from cwltool.errors import WorkflowException
 from cwltool.draft2tool import revmap_file, CommandLineTool
 from cwltool.load_tool import fetch_document
 from cwltool.builder import Builder
-from cwltool.pathmapper import adjustDirObjs
+from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
 
 from schema_salad.sourceline import SourceLine
 
 import ruamel.yaml as yaml
 
 import arvados.collection
+from arvados.errors import ApiError
 
 from .arvdocker import arv_docker_get_image
-from .runner import Runner, arvados_jobs_image, packed_workflow, trim_listing, upload_workflow_collection
-from .pathmapper import InitialWorkDirPathMapper
+from .runner import Runner, arvados_jobs_image, packed_workflow, upload_workflow_collection, trim_anonymous_location
+from .pathmapper import VwdPathMapper, trim_listing
 from .perf import Perf
 from . import done
 from ._version import __version__
@@ -51,8 +52,8 @@ class ArvadosJob(object):
                                                     keep_client=self.arvrunner.keep_client,
                                                     num_retries=self.arvrunner.num_retries)
                 script_parameters["task.vwd"] = {}
-                generatemapper = InitialWorkDirPathMapper([self.generatefiles], "", "",
-                                                          separateDirs=False)
+                generatemapper = VwdPathMapper([self.generatefiles], "", "",
+                                               separateDirs=False)
 
                 with Perf(metrics, "createfiles %s" % self.name):
                     for f, p in generatemapper.items():
@@ -60,8 +61,9 @@ class ArvadosJob(object):
                             with vwd.open(p.target, "w") as n:
                                 n.write(p.resolved.encode("utf-8"))
 
-                with Perf(metrics, "generatefiles.save_new %s" % self.name):
-                    vwd.save_new()
+                if vwd:
+                    with Perf(metrics, "generatefiles.save_new %s" % self.name):
+                        vwd.save_new()
 
                 for f, p in generatemapper.items():
                     if p.type == "File":
@@ -109,6 +111,7 @@ class ArvadosJob(object):
         if runtime_req:
             if "keep_cache" in runtime_req:
                 runtime_constraints["keep_cache_mb_per_task"] = runtime_req["keep_cache"]
+                runtime_constraints["min_ram_mb_per_node"] += runtime_req["keep_cache"]
             if "outputDirType" in runtime_req:
                 if runtime_req["outputDirType"] == "local_output_dir":
                     script_parameters["task.keepTmpOutput"] = False
@@ -121,6 +124,12 @@ class ArvadosJob(object):
         if not self.arvrunner.ignore_docker_for_reuse:
             filters.append(["docker_image_locator", "in docker", runtime_constraints["docker_image"]])
 
+        enable_reuse = kwargs.get("enable_reuse", True)
+        if enable_reuse:
+            reuse_req, _ = get_feature(self, "http://arvados.org/cwl#ReuseRequirement")
+            if reuse_req:
+                enable_reuse = reuse_req["enableReuse"]
+
         try:
             with Perf(metrics, "create %s" % self.name):
                 response = self.arvrunner.api.jobs().create(
@@ -134,15 +143,31 @@ class ArvadosJob(object):
                         "runtime_constraints": runtime_constraints
                     },
                     filters=filters,
-                    find_or_create=kwargs.get("enable_reuse", True)
+                    find_or_create=enable_reuse
                 ).execute(num_retries=self.arvrunner.num_retries)
 
             self.arvrunner.processes[response["uuid"]] = self
 
             self.update_pipeline_component(response)
 
-            if response["state"] in ("Complete", "Failed", "Cancelled"):
-                logger.info("%s reuse job %s", self.arvrunner.label(self), response["uuid"])
+            if response["state"] == "Complete":
+                logger.info("%s reused job %s", self.arvrunner.label(self), response["uuid"])
+                # Give read permission to the desired project on reused jobs
+                if response["owner_uuid"] != self.arvrunner.project_uuid:
+                    try:
+                        self.arvrunner.api.links().create(body={
+                            'link_class': 'permission',
+                            'name': 'can_read',
+                            'tail_uuid': self.arvrunner.project_uuid,
+                            'head_uuid': response["uuid"],
+                            }).execute(num_retries=self.arvrunner.num_retries)
+                    except ApiError as e:
+                        # The user might not have "manage" access on the job: log
+                        # a message and continue.
+                        logger.info("Creating read permission on job %s: %s",
+                                    response["uuid"],
+                                    e)
+
                 with Perf(metrics, "done %s" % self.name):
                     self.done(response)
             else:
@@ -253,6 +278,8 @@ class RunnerJob(Runner):
             self.job_order["cwl:tool"] = "%s/workflow.cwl#main" % wf_pdh
 
         adjustDirObjs(self.job_order, trim_listing)
+        adjustFileObjs(self.job_order, trim_anonymous_location)
+        adjustDirObjs(self.job_order, trim_anonymous_location)
 
         if self.output_name:
             self.job_order["arv:output_name"] = self.output_name