Fix: When restating prod, clear intervals across all related snapshots, not just promoted ones (#5274)

Author: erindru

sqlmesh/core/plan/common.py

@@ -1,6 +1,15 @@
 from __future__ import annotations
+import typing as t
+import logging
+from dataclasses import dataclass, field
 
-from sqlmesh.core.snapshot import Snapshot
+from sqlmesh.core.state_sync import StateReader
+from sqlmesh.core.snapshot import Snapshot, SnapshotId, SnapshotIdAndVersion, SnapshotNameVersion
+from sqlmesh.core.snapshot.definition import Interval
+from sqlmesh.utils.dag import DAG
+from sqlmesh.utils.date import now_timestamp
+
+logger = logging.getLogger(__name__)
 
 
 def should_force_rebuild(old: Snapshot, new: Snapshot) -> bool:
@@ -27,3 +36,176 @@ def is_breaking_kind_change(old: Snapshot, new: Snapshot) -> bool:
         # If the partitioning hasn't changed, then we don't need to rebuild
         return False
     return True
+
+
+@dataclass
+class SnapshotIntervalClearRequest:
+    # affected snapshot
+    snapshot: SnapshotIdAndVersion
+
+    # which interval to clear
+    interval: Interval
+
+    # which environments this snapshot is currently promoted
+    # note that this can be empty if the snapshot exists because its ttl has not expired
+    # but it is not part of any particular environment
+    environment_names: t.Set[str] = field(default_factory=set)
+
+    @property
+    def snapshot_id(self) -> SnapshotId:
+        return self.snapshot.snapshot_id
+
+    @property
+    def sorted_environment_names(self) -> t.List[str]:
+        return list(sorted(self.environment_names))
+
+
+def identify_restatement_intervals_across_snapshot_versions(
+    state_reader: StateReader,
+    prod_restatements: t.Dict[str, Interval],
+    disable_restatement_models: t.Set[str],
+    loaded_snapshots: t.Dict[SnapshotId, Snapshot],
+    current_ts: t.Optional[int] = None,
+) -> t.Dict[SnapshotId, SnapshotIntervalClearRequest]:
+    """
+    Given a map of snapshot names + intervals to restate in prod:
+        - Look up matching snapshots (match based on name - regardless of version, to get all versions)
+        - For each match, also match downstream snapshots in each dev environment while filtering out models that have restatement disabled
+        - Return a list of all snapshots that are affected + the interval that needs to be cleared for each
+
+    The goal here is to produce a list of intervals to invalidate across all dev snapshots so that a subsequent plan or
+    cadence run in those environments causes the intervals to be repopulated.
+    """
+    if not prod_restatements:
+        return {}
+
+    # Although :loaded_snapshots is sourced from RestatementStage.all_snapshots, since the only time we ever need
+    # to clear intervals across all environments is for prod, the :loaded_snapshots here are always from prod
+    prod_name_versions: t.Set[SnapshotNameVersion] = {
+        s.name_version for s in loaded_snapshots.values()
+    }
+
+    snapshot_intervals_to_clear: t.Dict[SnapshotId, SnapshotIntervalClearRequest] = {}
+
+    for env_summary in state_reader.get_environments_summary():
+        # Fetch the full environment object one at a time to avoid loading all environments into memory at once
+        env = state_reader.get_environment(env_summary.name)
+        if not env:
+            logger.warning("Environment %s not found", env_summary.name)
+            continue
+
+        snapshots_by_name = {s.name: s.table_info for s in env.snapshots}
+
+        # We dont just restate matching snapshots, we also have to restate anything downstream of them
+        # so that if A gets restated in prod and dev has A <- B <- C, B and C get restated in dev
+        env_dag = DAG({s.name: {p.name for p in s.parents} for s in env.snapshots})
+
+        for restate_snapshot_name, interval in prod_restatements.items():
+            if restate_snapshot_name not in snapshots_by_name:
+                # snapshot is not promoted in this environment
+                continue
+
+            affected_snapshot_names = [
+                x
+                for x in ([restate_snapshot_name] + env_dag.downstream(restate_snapshot_name))
+                if x not in disable_restatement_models
+            ]
+
+            for affected_snapshot_name in affected_snapshot_names:
+                affected_snapshot = snapshots_by_name[affected_snapshot_name]
+
+                # Don't clear intervals for a dev snapshot if it shares the same physical version with prod.
+                # Otherwise, prod will be affected by what should be a dev operation
+                if affected_snapshot.name_version in prod_name_versions:
+                    continue
+
+                clear_request = snapshot_intervals_to_clear.get(affected_snapshot.snapshot_id)
+                if not clear_request:
+                    clear_request = SnapshotIntervalClearRequest(
+                        snapshot=affected_snapshot.id_and_version, interval=interval
+                    )
+                    snapshot_intervals_to_clear[affected_snapshot.snapshot_id] = clear_request
+
+                clear_request.environment_names |= set([env.name])
+
+    # snapshot_intervals_to_clear now contains the entire hierarchy of affected snapshots based
+    # on building the DAG for each environment and including downstream snapshots
+    # but, what if there are affected snapshots that arent part of any environment?
+    unique_snapshot_names = set(snapshot_id.name for snapshot_id in snapshot_intervals_to_clear)
+
+    current_ts = current_ts or now_timestamp()
+    all_matching_non_prod_snapshots = {
+        s.snapshot_id: s
+        for s in state_reader.get_snapshots_by_names(
+            snapshot_names=unique_snapshot_names, current_ts=current_ts, exclude_expired=True
+        )
+        # Don't clear intervals for a snapshot if it shares the same physical version with prod.
+        # Otherwise, prod will be affected by what should be a dev operation
+        if s.name_version not in prod_name_versions
+    }
+
+    # identify the ones that we havent picked up yet, which are the ones that dont exist in any environment
+    if remaining_snapshot_ids := set(all_matching_non_prod_snapshots).difference(
+        snapshot_intervals_to_clear
+    ):
+        # these snapshot id's exist in isolation and may be related to a downstream dependency of the :prod_restatements,
+        # rather than directly related, so we can't simply look up the interval to clear based on :prod_restatements.
+        # To figure out the interval that should be cleared, we can match to the existing list based on name
+        # and conservatively take the widest interval that shows up
+        snapshot_name_to_widest_interval: t.Dict[str, Interval] = {}
+        for s_id, clear_request in snapshot_intervals_to_clear.items():
+            current_start, current_end = snapshot_name_to_widest_interval.get(
+                s_id.name, clear_request.interval
+            )
+            next_start, next_end = clear_request.interval
+
+            next_start = min(current_start, next_start)
+            next_end = max(current_end, next_end)
+
+            snapshot_name_to_widest_interval[s_id.name] = (next_start, next_end)
+
+        for remaining_snapshot_id in remaining_snapshot_ids:
+            remaining_snapshot = all_matching_non_prod_snapshots[remaining_snapshot_id]
+            snapshot_intervals_to_clear[remaining_snapshot_id] = SnapshotIntervalClearRequest(
+                snapshot=remaining_snapshot,
+                interval=snapshot_name_to_widest_interval[remaining_snapshot_id.name],
+            )
+
+    # for any affected full_history_restatement_only snapshots, we need to widen the intervals being restated to
+    # include the whole time range for that snapshot. This requires a call to state to load the full snapshot record,
+    # so we only do it if necessary
+    full_history_restatement_snapshot_ids = [
+        # FIXME: full_history_restatement_only is just one indicator that the snapshot can only be fully refreshed, the other one is Model.depends_on_self
+        # however, to figure out depends_on_self, we have to render all the model queries which, alongside having to fetch full snapshots from state,
+        # is problematic in secure environments that are deliberately isolated from arbitrary user code (since rendering a query may require user macros to be present)
+        # So for now, these are not considered
+        s_id
+        for s_id, s in snapshot_intervals_to_clear.items()
+        if s.snapshot.full_history_restatement_only
+    ]
+    if full_history_restatement_snapshot_ids:
+        # only load full snapshot records that we havent already loaded
+        additional_snapshots = state_reader.get_snapshots(
+            [
+                s.snapshot_id
+                for s in full_history_restatement_snapshot_ids
+                if s.snapshot_id not in loaded_snapshots
+            ]
+        )
+
+        all_snapshots = loaded_snapshots | additional_snapshots
+
+        for full_snapshot_id in full_history_restatement_snapshot_ids:
+            full_snapshot = all_snapshots[full_snapshot_id]
+            intervals_to_clear = snapshot_intervals_to_clear[full_snapshot_id]
+
+            original_start, original_end = intervals_to_clear.interval
+
+            # get_removal_interval() widens intervals if necessary
+            new_interval = full_snapshot.get_removal_interval(
+                start=original_start, end=original_end
+            )
+
+            intervals_to_clear.interval = new_interval
+
+    return snapshot_intervals_to_clear

sqlmesh/core/plan/evaluator.py

@@ -22,7 +22,7 @@
 from sqlmesh.core.console import Console, get_console
 from sqlmesh.core.environment import EnvironmentNamingInfo, execute_environment_statements
 from sqlmesh.core.macros import RuntimeStage
-from sqlmesh.core.snapshot.definition import Interval, to_view_mapping
+from sqlmesh.core.snapshot.definition import to_view_mapping
 from sqlmesh.core.plan import stages
 from sqlmesh.core.plan.definition import EvaluatablePlan
 from sqlmesh.core.scheduler import Scheduler
@@ -33,17 +33,15 @@
     SnapshotIntervals,
     SnapshotId,
     SnapshotInfoLike,
-    SnapshotTableInfo,
     SnapshotCreationFailedError,
-    SnapshotNameVersion,
 )
 from sqlmesh.utils import to_snake_case
 from sqlmesh.core.state_sync import StateSync
+from sqlmesh.core.plan.common import identify_restatement_intervals_across_snapshot_versions
 from sqlmesh.utils import CorrelationId
 from sqlmesh.utils.concurrency import NodeExecutionFailedError
 from sqlmesh.utils.errors import PlanError, SQLMeshError
-from sqlmesh.utils.dag import DAG
-from sqlmesh.utils.date import now
+from sqlmesh.utils.date import now, to_timestamp
 
 logger = logging.getLogger(__name__)
 
@@ -289,7 +287,9 @@ def visit_audit_only_run_stage(
     def visit_restatement_stage(
         self, stage: stages.RestatementStage, plan: EvaluatablePlan
     ) -> None:
-        snapshot_intervals_to_restate = {(s, i) for s, i in stage.snapshot_intervals.items()}
+        snapshot_intervals_to_restate = {
+            (s.id_and_version, i) for s, i in stage.snapshot_intervals.items()
+        }
 
         # Restating intervals on prod plans should mean that the intervals are cleared across
         # all environments, not just the version currently in prod
@@ -298,11 +298,16 @@ def visit_restatement_stage(
         #
         # Without this rule, its possible that promoting a dev table to prod will introduce old data to prod
         snapshot_intervals_to_restate.update(
-            self._restatement_intervals_across_all_environments(
-                prod_restatements=plan.restatements,
-                disable_restatement_models=plan.disabled_restatement_models,
-                loaded_snapshots={s.snapshot_id: s for s in stage.all_snapshots.values()},
-            )
+            {
+                (s.snapshot, s.interval)
+                for s in identify_restatement_intervals_across_snapshot_versions(
+                    state_reader=self.state_sync,
+                    prod_restatements=plan.restatements,
+                    disable_restatement_models=plan.disabled_restatement_models,
+                    loaded_snapshots={s.snapshot_id: s for s in stage.all_snapshots.values()},
+                    current_ts=to_timestamp(plan.execution_time or now()),
+                ).values()
+            }
         )
 
         self.state_sync.remove_intervals(
@@ -422,97 +427,6 @@ def _demote_snapshots(
             on_complete=on_complete,
         )
 
-    def _restatement_intervals_across_all_environments(
-        self,
-        prod_restatements: t.Dict[str, Interval],
-        disable_restatement_models: t.Set[str],
-        loaded_snapshots: t.Dict[SnapshotId, Snapshot],
-    ) -> t.Set[t.Tuple[SnapshotTableInfo, Interval]]:
-        """
-        Given a map of snapshot names + intervals to restate in prod:
-         - Look up matching snapshots across all environments (match based on name - regardless of version)
-         - For each match, also match downstream snapshots while filtering out models that have restatement disabled
-         - Return all matches mapped to the intervals of the prod snapshot being restated
-
-        The goal here is to produce a list of intervals to invalidate across all environments so that a cadence
-        run in those environments causes the intervals to be repopulated
-        """
-        if not prod_restatements:
-            return set()
-
-        prod_name_versions: t.Set[SnapshotNameVersion] = {
-            s.name_version for s in loaded_snapshots.values()
-        }
-
-        snapshots_to_restate: t.Dict[SnapshotId, t.Tuple[SnapshotTableInfo, Interval]] = {}
-
-        for env_summary in self.state_sync.get_environments_summary():
-            # Fetch the full environment object one at a time to avoid loading all environments into memory at once
-            env = self.state_sync.get_environment(env_summary.name)
-            if not env:
-                logger.warning("Environment %s not found", env_summary.name)
-                continue
-
-            keyed_snapshots = {s.name: s.table_info for s in env.snapshots}
-
-            # We dont just restate matching snapshots, we also have to restate anything downstream of them
-            # so that if A gets restated in prod and dev has A <- B <- C, B and C get restated in dev
-            env_dag = DAG({s.name: {p.name for p in s.parents} for s in env.snapshots})
-
-            for restatement, intervals in prod_restatements.items():
-                if restatement not in keyed_snapshots:
-                    continue
-                affected_snapshot_names = [
-                    x
-                    for x in ([restatement] + env_dag.downstream(restatement))
-                    if x not in disable_restatement_models
-                ]
-                snapshots_to_restate.update(
-                    {
-                        keyed_snapshots[a].snapshot_id: (keyed_snapshots[a], intervals)
-                        for a in affected_snapshot_names
-                        # Don't restate a snapshot if it shares the version with a snapshot in prod
-                        if keyed_snapshots[a].name_version not in prod_name_versions
-                    }
-                )
-
-        # for any affected full_history_restatement_only snapshots, we need to widen the intervals being restated to
-        # include the whole time range for that snapshot. This requires a call to state to load the full snapshot record,
-        # so we only do it if necessary
-        full_history_restatement_snapshot_ids = [
-            # FIXME: full_history_restatement_only is just one indicator that the snapshot can only be fully refreshed, the other one is Model.depends_on_self
-            # however, to figure out depends_on_self, we have to render all the model queries which, alongside having to fetch full snapshots from state,
-            # is problematic in secure environments that are deliberately isolated from arbitrary user code (since rendering a query may require user macros to be present)
-            # So for now, these are not considered
-            s_id
-            for s_id, s in snapshots_to_restate.items()
-            if s[0].full_history_restatement_only
-        ]
-        if full_history_restatement_snapshot_ids:
-            # only load full snapshot records that we havent already loaded
-            additional_snapshots = self.state_sync.get_snapshots(
-                [
-                    s.snapshot_id
-                    for s in full_history_restatement_snapshot_ids
-                    if s.snapshot_id not in loaded_snapshots
-                ]
-            )
-
-            all_snapshots = loaded_snapshots | additional_snapshots
-
-            for full_snapshot_id in full_history_restatement_snapshot_ids:
-                full_snapshot = all_snapshots[full_snapshot_id]
-                _, original_intervals = snapshots_to_restate[full_snapshot_id]
-                original_start, original_end = original_intervals
-
-                # get_removal_interval() widens intervals if necessary
-                new_intervals = full_snapshot.get_removal_interval(
-                    start=original_start, end=original_end
-                )
-                snapshots_to_restate[full_snapshot_id] = (full_snapshot.table_info, new_intervals)
-
-        return set(snapshots_to_restate.values())
-
     def _update_intervals_for_new_snapshots(self, snapshots: t.Collection[Snapshot]) -> None:
         snapshots_intervals: t.List[SnapshotIntervals] = []
         for snapshot in snapshots:

sqlmesh/core/snapshot/__init__.py

@@ -11,6 +11,7 @@
     SnapshotId as SnapshotId,
     SnapshotIdBatch as SnapshotIdBatch,
     SnapshotIdLike as SnapshotIdLike,
+    SnapshotIdAndVersionLike as SnapshotIdAndVersionLike,
     SnapshotInfoLike as SnapshotInfoLike,
     SnapshotIntervals as SnapshotIntervals,
     SnapshotNameVersion as SnapshotNameVersion,