[u a 8/8] Configure mirroring per source (#7066)

nadove-ucsc · achave11-ucsc · commit a4b4c1fc6943 · 2025-11-19T17:05:51.000-08:00
diff --git a/UPGRADING.rst b/UPGRADING.rst
@@ -20,6 +20,34 @@ reverted. This is all fairly informal and loosely defined. Hopefully we won't
 have too many entries in this file.
 
 
+#7066 Configure mirroring per catalog and source
+================================================
+
+In your personal deployments' ``environment.py`` file(s):
+
+1. Rename the loop variable ``internal`` to ``is_it``.
+
+2. Set the value of the ``pop`` flag to ``1``.
+
+3. Rename to type alias ``ProjectName`` to ``DatasetName``.
+
+4. In AnVIL deployments, rename the local variable ``project`` in the functions
+   ``mksrc`` and ``mkdelta`` to ``google_project`` and ``dataset`` respectively.
+
+5. Rename the following functions: ``bqsrc`` (AnVIL deployments only),
+   ``mksrc``, ``mkdelta``, ``mklist``, and ``mkdict``.
+
+6. Update the type annotations for those functions.
+
+7. Update the definitions of ``source`` and ``condense``.
+
+8. Insert the ``mirror_limit`` property in the definition of
+   ``AZUL_CATALOGS``.
+
+As always, use the sandbox deployment's ``environment.py`` as a model when
+upgrading personal deployments.
+
+
 #7468 Update Swagger to v5.29.4
 ===============================
 
diff --git a/deployments/anvilbox/environment.py b/deployments/anvilbox/environment.py
@@ -9,6 +9,7 @@
 is_sandbox = True
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -33,7 +34,7 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             flags: int = 0,
             ) -> tuple[DatasetName, SourceItem | None]:
     dataset = '_'.join(snapshot.split('_')[1:-3])
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -42,7 +43,9 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return dataset, source
 
diff --git a/deployments/anvildev/environment.py b/deployments/anvildev/environment.py
@@ -7,6 +7,7 @@
 )
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -31,7 +32,7 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             flags: int = 0,
             ) -> tuple[DatasetName, SourceItem | None]:
     dataset = '_'.join(snapshot.split('_')[1:-3])
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -40,7 +41,9 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return dataset, source
 
diff --git a/deployments/anvilprod/environment.py b/deployments/anvilprod/environment.py
@@ -9,6 +9,7 @@
 )
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -34,7 +35,7 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             flags: int = 0,
             ) -> tuple[DatasetName, SourceItem | None]:
     dataset = '_'.join(snapshot.split('_')[1:-3])
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -43,7 +44,9 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return dataset, source
 
diff --git a/deployments/dev/environment.py b/deployments/dev/environment.py
@@ -7,6 +7,7 @@
 )
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -20,7 +21,7 @@ def source(source_type: Literal['bigquery', 'parquet'],
            flags: int = 0,
            ) -> tuple[DatasetName, SourceItem | None]:
     _, env, project, _ = snapshot.split('_', 3)
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -29,7 +30,9 @@ def source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return project, source
 
diff --git a/deployments/hammerbox/environment.py b/deployments/hammerbox/environment.py
@@ -11,6 +11,7 @@
 is_sandbox = True
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -36,7 +37,7 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             flags: int = 0,
             ) -> tuple[DatasetName, SourceItem | None]:
     dataset = '_'.join(snapshot.split('_')[1:-3])
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -45,7 +46,9 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return dataset, source
 
diff --git a/deployments/prod/environment.py b/deployments/prod/environment.py
@@ -9,6 +9,7 @@
 )
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -22,7 +23,7 @@ def source(source_type: Literal['bigquery', 'parquet'],
            flags: int = 0,
            ) -> tuple[DatasetName, SourceItem | None]:
     _, env, project, _ = snapshot.split('_', 3)
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -31,7 +32,9 @@ def source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return project, source
 
diff --git a/deployments/sandbox/environment.py b/deployments/sandbox/environment.py
@@ -9,6 +9,7 @@
 is_sandbox = True
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -22,7 +23,7 @@ def source(source_type: Literal['bigquery', 'parquet'],
            flags: int = 0,
            ) -> tuple[DatasetName, SourceItem | None]:
     _, env, project, _ = snapshot.split('_', 3)
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -31,7 +32,9 @@ def source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return project, source
 
diff --git a/deployments/tempdev/environment.py b/deployments/tempdev/environment.py
@@ -7,6 +7,7 @@
 )
 
 pop = 1  # remove snapshot
+no_mirror = 2  # do not mirror files from snapshot (redundant for managed access snapshots)
 
 type DatasetName = str
 type SourceSpec = str
@@ -31,7 +32,7 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             flags: int = 0,
             ) -> tuple[DatasetName, SourceItem | None]:
     dataset = '_'.join(snapshot.split('_')[1:-3])
-    assert flags <= pop
+    assert flags <= pop | no_mirror
     source = None if flags & pop else (
         ':'.join([
             'tdr',
@@ -40,7 +41,9 @@ def _source(source_type: Literal['bigquery', 'parquet'],
             google_project,
             snapshot,
         ]),
-        {}
+        {
+            'mirror': not (flags & no_mirror),
+        }
     )
     return dataset, source
 
diff --git a/lambdas/service/app.py b/lambdas/service/app.py
@@ -121,7 +121,7 @@
         # changes and reset the minor version to zero. Otherwise, increment only
         # the minor version for backwards compatible changes. A backwards
         # compatible change is one that does not require updates to clients.
-        'version': '14.4',
+        'version': '14.5',
         'description': fd(f'''
             # Overview
 
diff --git a/lambdas/service/openapi.json b/lambdas/service/openapi.json
@@ -2,7 +2,7 @@
     "openapi": "3.0.1",
     "info": {
         "title": "azul-service-dev",
-        "version": "14.4",
+        "version": "14.5",
         "description": "\n# Overview\n\nAzul is a REST web service for querying metadata associated with\nboth experimental and analysis data from a data repository. In order\nto deliver response times that make it suitable for interactive use\ncases, the set of metadata properties that it exposes for sorting,\nfiltering, and aggregation is limited. Azul provides a uniform view\nof the metadata over a range of diverse schemas, effectively\nshielding clients from changes in the schemas as they occur over\ntime. It does so, however, at the expense of detail in the set of\nmetadata properties it exposes and in the accuracy with which it\naggregates them.\n\nAzul denormalizes and aggregates metadata into several different\nindices for selected entity types. Metadata entities can be queried\nusing the [Index](#operations-tag-Index) endpoints.\n\nA set of indices forms a catalog. There is a default catalog called\n`dcp2` which will be used unless a\ndifferent catalog name is specified using the `catalog` query\nparameter. Metadata from different catalogs is completely\nindependent: a response obtained by querying one catalog does not\nnecessarily correlate to a response obtained by querying another\none. Two catalogs can contain metadata from the same sources or\ndifferent sources. It is only guaranteed that the body of a\nresponse by any given endpoint adheres to one schema,\nindependently of which catalog was specified in the request.\n\nAzul provides the ability to download data and metadata via the\n[Manifests](#operations-tag-Manifests) endpoints. The\n`curl` format manifests can be used to\ndownload data files. Other formats provide various views of the\nmetadata. Manifests can be generated for a selection of files using\nfilters. These filters are interchangeable with the filters used by\nthe [Index](#operations-tag-Index) endpoints.\n\nAzul also provides a [summary](#operations-Index-get_index_summary)\nview of indexed data.\n\n## Data model\n\nAny index, when queried, returns a JSON array of hits. Each hit\nrepresents a metadata entity. Nested in each hit is a summary of the\nproperties of entities associated with the hit. An entity is\nassociated either by a direct edge in the original metadata graph,\nor indirectly as a series of edges. The nested properties are\ngrouped by the type of the associated entity. The properties of all\ndata files associated with a particular sample, for example, are\nlisted under `hits[*].files` in a `/index/samples` response. It is\nimportant to note that while each _hit_ represents a discrete\nentity, the properties nested within that hit are the result of an\naggregation over potentially many associated entities.\n\nTo illustrate this, consider a data file that is part of two\nprojects (a project is a group of related experiments, typically by\none laboratory, institution or consortium). Querying the `files`\nindex for this file yields a hit looking something like:\n\n```\n{\n    \"projects\": [\n        {\n            \"projectTitle\": \"Project One\"\n            \"laboratory\": ...,\n            ...\n        },\n        {\n            \"projectTitle\": \"Project Two\"\n            \"laboratory\": ...,\n            ...\n        }\n    ],\n    \"files\": [\n        {\n            \"format\": \"pdf\",\n            \"name\": \"Team description.pdf\",\n            ...\n        }\n    ]\n}\n```\n\nThis example hit contains two kinds of nested entities (a hit in an\nactual Azul response will contain more): There are the two projects\nentities, and the file itself. These nested entities contain\nselected metadata properties extracted in a consistent way. This\nmakes filtering and sorting simple.\n\nAlso notice that there is only one file. When querying a particular\nindex, the corresponding entity will always be a singleton like\nthis.\n\n\n## Contact us\n\nFor technical support please file an issue at\n[GitHub](https://github.com/DataBiosphere/azul/issues) or email\n`azul-group@ucsc.edu`. To report a security concern or misconduct please email\n`azul-group@ucsc.edu`.\n"
     },
     "tags": [
diff --git a/src/azul/azulclient.py b/src/azul/azulclient.py
@@ -401,9 +401,14 @@ def remote_mirror(self,
 
             def messages():
                 for source, cfg in sources:
-                    log.info('Mirroring files in source %r from catalog %r',
-                             str(source.spec), catalog)
-                    yield self.mirror_source_message(catalog, source)
+                    if cfg.mirror:
+                        log.info('Mirroring files in source %r from catalog %r',
+                                 str(source.spec), catalog)
+                        yield self.mirror_source_message(catalog, source)
+                    else:
+                        log.info('Not mirroring any files in source %r from catalog %r because '
+                                 'mirroring is explicitly disabled',
+                                 str(source.spec), catalog)
 
             self.queue_mirror_messages(messages())
 
diff --git a/src/azul/indexer/__init__.py b/src/azul/indexer/__init__.py
@@ -412,7 +412,7 @@ class SourceConfig(SerializableAttrs):
     """
     Configuration on how to index or mirror a specific source.
     """
-    pass
+    mirror: bool
 
 
 @attrs.frozen(kw_only=True, order=True)
diff --git a/test/azul_test_case.py b/test/azul_test_case.py
@@ -456,7 +456,7 @@ def catalog_config(cls) -> dict[CatalogName, Config.Catalog]:
                                         mirror_limit=None,
                                         plugins=dict(metadata=config.Catalog.Plugin(name='hca'),
                                                      repository=config.Catalog.Plugin(name='dss')),
-                                        sources={str(cls.source.spec): {}})
+                                        sources={str(cls.source.spec): {'mirror': True}})
         }
 
 
@@ -480,7 +480,7 @@ def _patch_tdr_service_url(cls):
 
     @classmethod
     def _sources(cls):
-        return {str(cls.source.spec): {}}
+        return {str(cls.source.spec): {'mirror': True}}
 
     @classmethod
     def _patch_source_cache(cls):
@@ -522,5 +522,5 @@ def catalog_config(cls) -> dict[CatalogName, Config.Catalog]:
                                         mirror_limit=None,
                                         plugins=dict(metadata=config.Catalog.Plugin(name='anvil'),
                                                      repository=config.Catalog.Plugin(name='tdr_anvil')),
-                                        sources={str(cls.source.spec): {}})
+                                        sources={str(cls.source.spec): {'mirror': True}})
         }
diff --git a/test/indexer/test_mirror_controller.py b/test/indexer/test_mirror_controller.py
@@ -130,8 +130,9 @@ def mirror_controller(self) -> MirrorController:
     def _mirror_event(self, body: JSON) -> list[SQSRecord]:
         return [self._mock_sqs_record(body, fifo=True)]
 
-    def _remote_mirror(self) -> MutableJSONs:
-        self.client.remote_mirror(self.catalog, [(self.source, SourceConfig())])
+    def _remote_mirror(self, mirror_source_cfg: bool = True) -> MutableJSONs:
+        cfg = SourceConfig(mirror=mirror_source_cfg)
+        self.client.remote_mirror(self.catalog, [(self.source, cfg)])
         return self._read_queue(self.client.mirror_queue())
 
     def _test_remote_mirror(self):
@@ -208,6 +209,11 @@ def test_info_schema(self):
 
     def test_files_not_mirrored(self):
         self._create_mock_queues(config.mirror_queue_names)
+
+        with self.subTest(no_mirror=True):
+            messages = self._remote_mirror(mirror_source_cfg=False)
+            self.assertEqual([], messages)
+
         catalog = config.catalogs[self.catalog]
 
         def patch_max_file_size(size):
diff --git a/test/integration_test.py b/test/integration_test.py
diff --git a/test/service/test_repository_files.py b/test/service/test_repository_files.py
