11002: Merge branch 'master' into 11002-arvput-crash-fix

diff --git a/sdk/python/arvados/commands/put.py b/sdk/python/arvados/commands/put.py
index c26bb04f3200778bb1b06d033ae958ece01ba731..30ba88256357059d3020cc5b894a11bef2e4bc5e 100644 (file)
--- a/sdk/python/arvados/commands/put.py
+++ b/sdk/python/arvados/commands/put.py
@@ -452,11 +452,15 @@ class ArvPutUploadJob(object):
                 # Stop the thread before doing anything else
                 self._stop_checkpointer.set()
                 self._checkpointer.join()
-                # Commit all pending blocks & one last _update()
-                self._local_collection.manifest_text()
-                self._update(final=True)
-                if save_collection:
-                    self.save_collection()
+                try:
+                    # Commit all pending blocks & one last _update()
+                    self._local_collection.manifest_text()
+                    self._update(final=True)
+                    if save_collection:
+                        self.save_collection()
+                except AttributeError:
+                    # Exception caught in inconsistent state, finish as is.
+                    self.logger.warning("Couldn't save last checkpoint while exiting.")
             if self.use_cache:
                 self._cache_file.close()
 

diff --git a/sdk/python/tests/test_arv_put.py b/sdk/python/tests/test_arv_put.py
index f1dfd03def33d09c1ede560f50c5a059020f9c0c..f9af01d9e85ab1266930974d95f6bd0a312ef52f 100644 (file)
--- a/sdk/python/tests/test_arv_put.py
+++ b/sdk/python/tests/test_arv_put.py
@@ -261,12 +261,19 @@ class ArvPutUploadJobTest(run_test_server.TestCaseWithServers,
             data = random.choice(['x', 'y', 'z']) * 1024 * 1024 # 1 MB
             fileobj.write(data)
         fileobj.close()
+        # Temp dir containing small files to be repacked
+        self.small_files_dir = tempfile.mkdtemp()
+        data = 'y' * 1024 * 1024 # 1 MB
+        for i in range(1, 70):
+            with open(os.path.join(self.small_files_dir, str(i)), 'w') as f:
+                f.write(data + str(i))
         self.arvfile_write = getattr(arvados.arvfile.ArvadosFileWriter, 'write')
 
     def tearDown(self):
         super(ArvPutUploadJobTest, self).tearDown()
         shutil.rmtree(self.tempdir)
         os.unlink(self.large_file_name)
+        shutil.rmtree(self.small_files_dir)
 
     def test_writer_works_without_cache(self):
         cwriter = arv_put.ArvPutUploadJob(['/dev/null'], resume=False)
@@ -337,6 +344,27 @@ class ArvPutUploadJobTest(run_test_server.TestCaseWithServers,
                          os.path.getsize(self.large_file_name))
         writer2.destroy_cache()
 
+    # Test for bug #11002
+    def test_graceful_exit_while_repacking_small_blocks(self):
+        def wrapped_commit(*args, **kwargs):
+            raise SystemExit("Simulated error")
+
+        with mock.patch('arvados.arvfile._BlockManager.commit_bufferblock',
+                        autospec=True) as mocked_commit:
+            mocked_commit.side_effect = wrapped_commit
+            # Upload a little more than 1 block, wrapped_commit will make the first block
+            # commit to fail.
+            # arv-put should not exit with an exception by trying to commit the collection
+            # as it's in an inconsistent state.
+            writer = arv_put.ArvPutUploadJob([self.small_files_dir],
+                                             replication_desired=1)
+            try:
+                with self.assertRaises(SystemExit):
+                    writer.start(save_collection=False)
+            except AttributeError:
+                self.fail("arv-put command is trying to use a corrupted BlockManager. See https://dev.arvados.org/issues/11002")
+        writer.destroy_cache()
+
     def test_no_resume_when_asked(self):
         def wrapped_write(*args, **kwargs):
             data = args[1]