From e9c0ab574e6bae899556e82d4ba0da439a976093 Mon Sep 17 00:00:00 2001
From: Peter Amstutz <peter.amstutz@curoverse.com>
Date: Mon, 7 Jul 2014 08:59:24 -0400
Subject: [PATCH] run-command retry after exception when writing output to
 keep.  no issue #

---
 crunch_scripts/run-command | 24 +++++++++++++++++-------
 1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/crunch_scripts/run-command b/crunch_scripts/run-command
index 6335523a74..a8aae24f16 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")
@@ -71,13 +74,20 @@ 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("..")
-- 
2.39.5