Change PUT to fail on tracker DB errors

The original goal is to create a Dataset only on dataset intake;
all other operations will require that a Dataset already exist
and fail in some manner if not. This is the first step towards
that, with a server PUT upload attempt failing (not storing the
uploaded dataset) if we're unable to create a Dataset.

Author: dbutenhof

lib/pbench/server/api/resources/upload_api.py

@@ -3,6 +3,8 @@
 import tempfile
 import hashlib
 from pathlib import Path
+from http import HTTPStatus
+
 from flask_restful import Resource, abort
 from flask import request, jsonify
 from werkzeug.utils import secure_filename
@@ -114,8 +116,16 @@ def put(self, controller):
         # TODO: Need real user from PUT!
 
         # Create a tracking dataset object; it'll begin in UPLOADING state
-        dataset = Dataset.create(controller=controller, path=tar_full_path, md5=md5sum)
-        self.logger.info("Uploading {}", dataset)
+        try:
+            dataset = Dataset(controller=controller, path=tar_full_path, md5=md5sum)
+            dataset.add()
+        except Exception:
+            self.logger.exception("unable to create dataset for {}", filename)
+            abort(
+                HTTPStatus.INTERNAL_SERVER_ERROR, message="INTERNAL ERROR",
+            )
+
+        self.logger.info("Uploading file {} to {}", filename, dataset)
 
         with tempfile.NamedTemporaryFile(mode="wb", dir=path) as ofp:
             chunk_size = 4096
@@ -189,11 +199,11 @@ def put(self, controller):
                 )
                 raise
 
-        if dataset:
-            try:
-                dataset.advance(States.UPLOADED)
-            except Exception:
-                self.logger.exception("Unable to finalize {}", dataset)
+        try:
+            dataset.advance(States.UPLOADED)
+        except Exception:
+            self.logger.exception("Unable to finalize {}", dataset)
+            abort(HTTPStatus.INTERNAL_SERVER_ERROR, message="INTERNAL ERROR")
         response = jsonify(dict(message="File successfully uploaded"))
         response.status_code = 201
         return response

lib/pbench/server/db/models/tracker.py

@@ -239,13 +239,10 @@ class States(enum.Enum):
 
     UPLOADING = ("Uploading", True)
     UPLOADED = ("Uploaded", False)
-    ARCHIVING = ("Archiving", True)
-    ARCHIVED = ("Archived", False)
     UNPACKING = ("Unpacking", True)
     UNPACKED = ("Unpacked", False)
     INDEXING = ("Indexing", True)
     INDEXED = ("Indexed", False)
-    REINDEX = ("Re-index", False)
     EXPIRING = ("Expiring", True)
     EXPIRED = ("Expired", False)
     QUARANTINED = ("Quarantined", False)
@@ -302,7 +299,7 @@ class Dataset(Database.Base):
         # because they're terminal states that cannot be exited.
     }
 
-    DEFAULT_USERNAME = "public"
+    ANTI_USER = "public"  # The "user" that owns unowned datasets
 
     id = Column(Integer, primary_key=True, autoincrement=True)
     user = Column(String(255), unique=False, nullable=False)
@@ -367,13 +364,13 @@ def __init__(self, **kwargs):
             if type(kwargs.get("state")) is not States:
                 raise DatasetBadParameterType(kwargs.get("state"), States)
         super().__init__(**kwargs)
-        self.user = kwargs.get("user") if "user" in kwargs else Dataset.DEFAULT_USERNAME
+        self.user = kwargs.get("user", Dataset.ANTI_USER)
         self.controller = kwargs.get("controller")
         self.name = kwargs.get("name")
         time = datetime.datetime.now()
-        self.created = kwargs.get("created") if "created" in kwargs else time
-        self.md5 = kwargs.get("md5") if "md5" in kwargs else None
-        self.state = kwargs.get("state") if "state" in kwargs else States.UPLOADING
+        self.created = time
+        self.md5 = kwargs.get("md5")
+        self.state = kwargs.get("state", States.UPLOADING)
         self.transition = time
 
     @staticmethod

lib/pbench/test/unit/server/test_state_tracker.py

@@ -17,7 +17,7 @@ class TestStateTracker:
     def test_state_enum(self):
         """ Test the States ENUM properties
         """
-        assert len(States.__members__) == 12
+        assert len(States.__members__) == 9
         for n, s in States.__members__.items():
             assert str(s) == s.friendly
             assert s.mutating == (
@@ -66,7 +66,7 @@ def test_attach_none(self):
         does not exist.
         """
         with pytest.raises(DatasetNotFound):
-            Dataset.attach(controller="frodo", name="venus", state=States.ARCHIVING)
+            Dataset.attach(controller="frodo", name="venus", state=States.UPLOADING)
 
     def test_attach_controller_path(self):
         """ Test that we can attach using controller and name to a