diff --git a/swh/storage/cassandra/cql.py b/swh/storage/cassandra/cql.py
--- a/swh/storage/cassandra/cql.py
+++ b/swh/storage/cassandra/cql.py
@@ -243,8 +243,6 @@
     # 'content' table
     ##########################
 
-    _content_pk = ["sha1", "sha1_git", "sha256", "blake2s256"]
-
     def _content_add_finalize(self, statement: BoundStatement) -> None:
         """Returned currified by content_add_prepare, to be called when the
         content row should be added to the primary table."""
@@ -265,7 +263,7 @@
         token_class = self._cluster.metadata.token_map.token_class
 
         # Token of the row when it will be inserted. This is equivalent to
-        # "SELECT token({', '.join(self._content_pk)}) FROM content WHERE ..."
+        # "SELECT token({', '.join(ContentRow.PARTITION_KEY)}) FROM content WHERE ..."
         # after the row is inserted; but we need the token to insert in the
         # index tables *before* inserting to the main 'content' table
         token = token_class.from_key(statement.routing_key).value
@@ -295,13 +293,13 @@
             return None
 
     @_prepared_select_statement(
-        ContentRow, f"WHERE token({', '.join(_content_pk)}) = ?"
+        ContentRow, f"WHERE token({', '.join(ContentRow.PARTITION_KEY)}) = ?"
     )
     def content_get_from_token(self, token, *, statement) -> Iterable[ContentRow]:
         return map(ContentRow.from_dict, self._execute_with_retries(statement, [token]))
 
     @_prepared_select_statement(
-        ContentRow, f"WHERE token({', '.join(_content_pk)}) > ? LIMIT 1"
+        ContentRow, f"WHERE token({', '.join(ContentRow.PARTITION_KEY)}) > ? LIMIT 1"
     )
     def content_get_random(self, *, statement) -> Optional[ContentRow]:
         return self._get_random_row(ContentRow, statement)
@@ -310,7 +308,7 @@
         (
             "SELECT token({0}) AS tok, {1} FROM content "
             "WHERE token({0}) >= ? AND token({0}) <= ? LIMIT ?"
-        ).format(", ".join(_content_pk), ", ".join(ContentRow.cols()))
+        ).format(", ".join(ContentRow.PARTITION_KEY), ", ".join(ContentRow.cols()))
     )
     def content_get_token_range(
         self, start: int, end: int, limit: int, *, statement
@@ -354,7 +352,6 @@
     # 'skipped_content' table
     ##########################
 
-    _skipped_content_pk = ["sha1", "sha1_git", "sha256", "blake2s256"]
     _magic_null_pk = b"<null>"
     """
     NULLs (or all-empty blobs) are not allowed in primary keys; instead use a
@@ -377,7 +374,7 @@
 
         # Replace NULLs (which are not allowed in the partition key) with
         # an empty byte string
-        for key in self._skipped_content_pk:
+        for key in SkippedContentRow.PARTITION_KEY:
             if getattr(content, key) is None:
                 setattr(content, key, self._magic_null_pk)
 
@@ -388,7 +385,7 @@
         token_class = self._cluster.metadata.token_map.token_class
 
         # Token of the row when it will be inserted. This is equivalent to
-        # "SELECT token({', '.join(self._content_pk)})
+        # "SELECT token({', '.join(SkippedContentRow.PARTITION_KEY)})
         #  FROM skipped_content WHERE ..."
         # after the row is inserted; but we need the token to insert in the
         # index tables *before* inserting to the main 'skipped_content' table
diff --git a/swh/storage/cassandra/model.py b/swh/storage/cassandra/model.py
--- a/swh/storage/cassandra/model.py
+++ b/swh/storage/cassandra/model.py
@@ -21,7 +21,7 @@
 
 import dataclasses
 import datetime
-from typing import Any, ClassVar, Dict, List, Optional, Type, TypeVar
+from typing import Any, ClassVar, Dict, List, Optional, Tuple, Type, TypeVar
 
 from swh.model.model import Person, TimestampWithTimezone
 
@@ -31,6 +31,8 @@
 
 class BaseRow:
     TABLE: ClassVar[str]
+    PARTITION_KEY: ClassVar[Tuple[str, ...]]
+    CLUSTERING_KEY: ClassVar[Tuple[str, ...]] = ()
 
     @classmethod
     def from_dict(cls: Type[T], d: Dict[str, Any]) -> T:
@@ -47,6 +49,7 @@
 @dataclasses.dataclass
 class ContentRow(BaseRow):
     TABLE = "content"
+    PARTITION_KEY = ("sha1", "sha1_git", "sha256", "blake2s256")
 
     sha1: bytes
     sha1_git: bytes
@@ -60,6 +63,7 @@
 @dataclasses.dataclass
 class SkippedContentRow(BaseRow):
     TABLE = "skipped_content"
+    PARTITION_KEY = ("sha1", "sha1_git", "sha256", "blake2s256")
 
     sha1: Optional[bytes]
     sha1_git: Optional[bytes]
@@ -75,6 +79,7 @@
 @dataclasses.dataclass
 class DirectoryRow(BaseRow):
     TABLE = "directory"
+    PARTITION_KEY = ("id",)
 
     id: bytes
 
@@ -82,6 +87,8 @@
 @dataclasses.dataclass
 class DirectoryEntryRow(BaseRow):
     TABLE = "directory_entry"
+    PARTITION_KEY = ("directory_id",)
+    CLUSTERING_KEY = ("name",)
 
     directory_id: bytes
     name: bytes
@@ -93,6 +100,7 @@
 @dataclasses.dataclass
 class RevisionRow(BaseRow):
     TABLE = "revision"
+    PARTITION_KEY = ("id",)
 
     id: bytes
     date: Optional[TimestampWithTimezone]
@@ -110,6 +118,8 @@
 @dataclasses.dataclass
 class RevisionParentRow(BaseRow):
     TABLE = "revision_parent"
+    PARTITION_KEY = ("id",)
+    CLUSTERING_KEY = ("parent_rank",)
 
     id: bytes
     parent_rank: int
@@ -119,6 +129,7 @@
 @dataclasses.dataclass
 class ReleaseRow(BaseRow):
     TABLE = "release"
+    PARTITION_KEY = ("id",)
 
     id: bytes
     target_type: str
@@ -133,6 +144,7 @@
 @dataclasses.dataclass
 class SnapshotRow(BaseRow):
     TABLE = "snapshot"
+    PARTITION_KEY = ("id",)
 
     id: bytes
 
@@ -140,6 +152,8 @@
 @dataclasses.dataclass
 class SnapshotBranchRow(BaseRow):
     TABLE = "snapshot_branch"
+    PARTITION_KEY = ("snapshot_id",)
+    CLUSTERING_KEY = ("name",)
 
     snapshot_id: bytes
     name: bytes
@@ -150,6 +164,8 @@
 @dataclasses.dataclass
 class OriginVisitRow(BaseRow):
     TABLE = "origin_visit"
+    PARTITION_KEY = ("origin",)
+    CLUSTERING_KEY = ("visit",)
 
     origin: str
     visit: int
@@ -160,6 +176,8 @@
 @dataclasses.dataclass
 class OriginVisitStatusRow(BaseRow):
     TABLE = "origin_visit_status"
+    PARTITION_KEY = ("origin",)
+    CLUSTERING_KEY = ("visit", "date")
 
     origin: str
     visit: int
@@ -172,6 +190,7 @@
 @dataclasses.dataclass
 class OriginRow(BaseRow):
     TABLE = "origin"
+    PARTITION_KEY = ("sha1",)
 
     sha1: bytes
     url: str
@@ -181,6 +200,8 @@
 @dataclasses.dataclass
 class MetadataAuthorityRow(BaseRow):
     TABLE = "metadata_authority"
+    PARTITION_KEY = ("url",)
+    CLUSTERING_KEY = ("type",)
 
     url: str
     type: str
@@ -190,6 +211,8 @@
 @dataclasses.dataclass
 class MetadataFetcherRow(BaseRow):
     TABLE = "metadata_fetcher"
+    PARTITION_KEY = ("name",)
+    CLUSTERING_KEY = ("version",)
 
     name: str
     version: str
@@ -199,6 +222,14 @@
 @dataclasses.dataclass
 class RawExtrinsicMetadataRow(BaseRow):
     TABLE = "raw_extrinsic_metadata"
+    PARTITION_KEY = ("id",)
+    CLUSTERING_KEY = (
+        "authority_type",
+        "authority_url",
+        "discovery_date",
+        "fetcher_name",
+        "fetcher_version",
+    )
 
     type: str
     id: str
@@ -224,6 +255,8 @@
 @dataclasses.dataclass
 class ObjectCountRow(BaseRow):
     TABLE = "object_count"
+    PARTITION_KEY = ("partition_key",)
+    CLUSTERING_KEY = ("object_type",)
 
     partition_key: int
     object_type: str