X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/b7de0ac72fe6e5270d58940f782e630789391f05..1875ddb761e4ae2909d2afe0718f3d0ad5f3ce0f:/crunch_scripts/run-command?ds=sidebyside

diff --git a/crunch_scripts/run-command b/crunch_scripts/run-command
index 6335523a74..2eb5589471 100755
--- a/crunch_scripts/run-command
+++ b/crunch_scripts/run-command
@@ -7,11 +7,14 @@ import subprocess
 import sys
 import shutil
 import subst
+import time
 
 os.umask(0077)
 
 t = arvados.current_task().tmpdir
 
+api = arvados.api('v1')
+
 os.chdir(arvados.current_task().tmpdir)
 os.mkdir("tmpdir")
 os.mkdir("output")
@@ -34,8 +37,12 @@ def sub_link(v):
 def sub_tmpdir(v):
     return os.path.join(arvados.current_task().tmpdir, 'tmpdir')
 
+def sub_cores(v):
+     return os.environ['CRUNCH_NODE_SLOTS']
+
 subst.default_subs["link "] = sub_link
 subst.default_subs["tmpdir"] = sub_tmpdir
+subst.default_subs["node.cores"] = sub_cores
 
 rcode = 1
 
@@ -71,13 +78,20 @@ finally:
 
     out = arvados.CollectionWriter()
     out.write_directory_tree(".", max_manifest_depth=0)
-    outuuid = out.finish()
-    arvados.api('v1').job_tasks().update(uuid=arvados.current_task()['uuid'],
-                                         body={
-                                             'output':outuuid,
-                                             'success': (rcode == 0),
-                                             'progress':1.0
-                                         }).execute()
+    done = False
+    while not done:
+        try:
+            outuuid = out.finish()
+            api.job_tasks().update(uuid=arvados.current_task()['uuid'],
+                                                 body={
+                                                     'output':outuuid,
+                                                     'success': (rcode == 0),
+                                                     'progress':1.0
+                                                 }).execute()
+            done = True
+        except Exception as e:
+            print("run-command: caught exception: {}".format(e))
+            time.sleep(5)
 
 if rcode == 0:
     os.chdir("..")