mgr/balancer: optimize 'balancer status detail'

Before, we were updating the balancer status by
iterating through all pg upmap entires. This was
affecting the loading time of other mgr modules
on clusters with a large number of pgs (600+).

This can be optimized by simply pulling from
the incremental.

Fixes: https://tracker.ceph.com/issues/68657
Signed-off-by: Laura Flores <[email protected]>

Author: Laura Flores

src/pybind/mgr/balancer/module.py

@@ -339,12 +339,10 @@ class Module(MgrModule):
     no_optimization_needed = False
     success_string = 'Optimization plan created successfully'
     in_progress_string = 'in progress'
-    last_pg_upmap: List[Dict[str, Any]] = []
     pg_upmap_items_added: List[Dict[str, Any]] = []
     pg_upmap_items_removed: List[Dict[str, Any]] = []
-    last_pg_upmap_primaries: List[Dict[str, Any]] = []
     pg_upmap_primaries_added: List[Dict[str, Any]] = []
-    pg_upmap_activity_initalized = False
+    pg_upmap_primaries_removed: List[Dict[str, Any]] = []
 
     def __init__(self, *args: Any, **kwargs: Any) -> None:
         super(Module, self).__init__(*args, **kwargs)
@@ -665,7 +663,7 @@ def plan_execute(self, plan: str) -> Tuple[int, str, str]:
         if not plan_:
             return (-errno.ENOENT, '', f'plan {plan} not found')
         r, detail = self.execute(plan_)
-        self.update_pg_upmap_activity()  # update pg activity in `balancer status detail`
+        self.update_pg_upmap_activity(plan_)  # update pg activity in `balancer status detail`
         self.plan_rm(plan)
         return (r, '', detail)
 
@@ -757,7 +755,7 @@ def serve(self) -> None:
                     self.execute(plan)
                 else:
                     self.optimize_result = detail
-                self.update_pg_upmap_activity()  # update pg activity in `balancer status detail`
+                self.update_pg_upmap_activity(plan)  # update pg activity in `balancer status detail`
                 self.optimizing = False
             self.log.debug('Sleeping for %d', sleep_interval)
             self.event.wait(sleep_interval)
@@ -1582,22 +1580,16 @@ def gather_telemetry(self) -> Dict[str, Any]:
             'mode': self.mode,
         }
 
-    def update_pg_upmap_activity(self) -> None:
-        osdmap = self.get_osdmap()
-        if not self.pg_upmap_activity_initalized:
-            self.last_pg_upmap = osdmap.dump().get('pg_upmap_items', '')
-            self.last_pg_upmap_primaries = osdmap.dump().get('pg_upmap_primaries', '')
-            self.pg_upmap_activity_initalized = True
+    def update_pg_upmap_activity(self, plan: Plan) -> None:
+        incdump = plan.inc.dump()
 
         # update pg_upmap_items
-        self.pg_upmap_items_added = [pg for pg in osdmap.dump().get('pg_upmap_items', '') if pg not in self.last_pg_upmap]
-        self.pg_upmap_items_removed = [pg for pg in self.last_pg_upmap if pg not in osdmap.dump().get('pg_upmap_items', '')]
-        self.last_pg_upmap = osdmap.dump().get('pg_upmap_items', '')
+        self.pg_upmap_items_added = incdump.get('new_pg_upmap_items', [])
+        self.pg_upmap_items_removed = incdump.get('old_pg_upmap_items', [])
 
         # update pg_upmap_primaries
-        self.pg_upmap_primaries_added = [pg for pg in osdmap.dump().get('pg_upmap_primaries', '') if pg not in self.last_pg_upmap_primaries]
-        self.pg_upmap_primaries_removed = [pg for pg in self.last_pg_upmap_primaries if pg not in osdmap.dump().get('pg_upmap_primaries', '')]
-        self.last_pg_upmap_primaries = osdmap.dump().get('pg_upmap_primaries', '')
+        self.pg_upmap_primaries_added = incdump.get('new_pg_upmap_primaries', [])
+        self.pg_upmap_primaries_removed = incdump.get('old_pg_upmap_primaries', [])
 
     def self_test(self) -> None:
         # turn balancer on