From: Ilya Dryomov <idryomov@gmail.com>
Date: Fri, 17 Jun 2022 08:28:55 +0000 (+0200)
Subject: mgr/rbd_support: refresh schedule queue immediately after delay elapses
X-Git-Tag: v18.0.0~653^2~1
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=ef3edd399adc99f3ff2acf580727d3dd5439d862;p=ceph.git

mgr/rbd_support: refresh schedule queue immediately after delay elapses

The existing logic often leads to refresh_pools() and refresh_images()
being invoked after a 120 second delay instead of after an intended 60
second delay.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
---

diff --git a/src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py b/src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py
index 013b210266bd9..ffcd357d2ed17 100644
--- a/src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py
+++ b/src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py
@@ -446,6 +446,7 @@ class MirrorSnapshotScheduleHandler:
     MODULE_OPTION_NAME = "mirror_snapshot_schedule"
     MODULE_OPTION_NAME_MAX_CONCURRENT_SNAP_CREATE = "max_concurrent_snap_create"
     SCHEDULE_OID = "rbd_mirror_snapshot_schedule"
+    REFRESH_DELAY_SECONDS = 60.0
 
     lock = Lock()
     condition = Condition(lock)
@@ -470,11 +471,11 @@ class MirrorSnapshotScheduleHandler:
         try:
             self.log.info("MirrorSnapshotScheduleHandler: starting")
             while True:
-                self.refresh_images()
+                refresh_delay = self.refresh_images()
                 with self.lock:
                     (image_spec, wait_time) = self.dequeue()
                     if not image_spec:
-                        self.condition.wait(min(wait_time, 60))
+                        self.condition.wait(min(wait_time, refresh_delay))
                         continue
                 pool_id, namespace, image_id = image_spec
                 self.create_snapshot_requests.add(pool_id, namespace, image_id)
@@ -502,9 +503,10 @@ class MirrorSnapshotScheduleHandler:
         with self.lock:
             self.schedules = schedules
 
-    def refresh_images(self) -> None:
-        if (datetime.now() - self.last_refresh_images).seconds < 60:
-            return
+    def refresh_images(self) -> float:
+        elapsed = (datetime.now() - self.last_refresh_images).total_seconds()
+        if elapsed < self.REFRESH_DELAY_SECONDS:
+            return self.REFRESH_DELAY_SECONDS - elapsed
 
         self.log.debug("MirrorSnapshotScheduleHandler: refresh_images")
 
@@ -517,7 +519,7 @@ class MirrorSnapshotScheduleHandler:
                 self.images = {}
                 self.queue = {}
                 self.last_refresh_images = datetime.now()
-                return
+                return self.REFRESH_DELAY_SECONDS
 
         epoch = int(datetime.now().strftime('%s'))
         images: Dict[str, Dict[str, Dict[str, str]]] = {}
@@ -535,6 +537,7 @@ class MirrorSnapshotScheduleHandler:
 
         self.watchers.unregister_stale(epoch)
         self.last_refresh_images = datetime.now()
+        return self.REFRESH_DELAY_SECONDS
 
     def load_pool_images(self,
                          ioctx: rados.Ioctx,
diff --git a/src/pybind/mgr/rbd_support/trash_purge_schedule.py b/src/pybind/mgr/rbd_support/trash_purge_schedule.py
index 3b4d906837617..1cb87d8a82cf0 100644
--- a/src/pybind/mgr/rbd_support/trash_purge_schedule.py
+++ b/src/pybind/mgr/rbd_support/trash_purge_schedule.py
@@ -16,6 +16,7 @@ from .schedule import LevelSpec, Interval, StartTime, Schedule, Schedules
 class TrashPurgeScheduleHandler:
     MODULE_OPTION_NAME = "trash_purge_schedule"
     SCHEDULE_OID = "rbd_trash_purge_schedule"
+    REFRESH_DELAY_SECONDS = 60.0
 
     lock = Lock()
     condition = Condition(lock)
@@ -35,11 +36,11 @@ class TrashPurgeScheduleHandler:
         try:
             self.log.info("TrashPurgeScheduleHandler: starting")
             while True:
-                self.refresh_pools()
+                refresh_delay = self.refresh_pools()
                 with self.lock:
                     (ns_spec, wait_time) = self.dequeue()
                     if not ns_spec:
-                        self.condition.wait(min(wait_time, 60))
+                        self.condition.wait(min(wait_time, refresh_delay))
                         continue
                 pool_id, namespace = ns_spec
                 self.trash_purge(pool_id, namespace)
@@ -74,9 +75,10 @@ class TrashPurgeScheduleHandler:
         with self.lock:
             self.schedules = schedules
 
-    def refresh_pools(self) -> None:
-        if (datetime.now() - self.last_refresh_pools).seconds < 60:
-            return
+    def refresh_pools(self) -> float:
+        elapsed = (datetime.now() - self.last_refresh_pools).total_seconds()
+        if elapsed < self.REFRESH_DELAY_SECONDS:
+            return self.REFRESH_DELAY_SECONDS - elapsed
 
         self.log.debug("TrashPurgeScheduleHandler: refresh_pools")
 
@@ -88,7 +90,7 @@ class TrashPurgeScheduleHandler:
                 self.pools = {}
                 self.queue = {}
                 self.last_refresh_pools = datetime.now()
-                return
+                return self.REFRESH_DELAY_SECONDS
 
         pools: Dict[str, Dict[str, str]] = {}
 
@@ -104,6 +106,7 @@ class TrashPurgeScheduleHandler:
             self.pools = pools
 
         self.last_refresh_pools = datetime.now()
+        return self.REFRESH_DELAY_SECONDS
 
     def load_pool(self, ioctx: rados.Ioctx, pools: Dict[str, Dict[str, str]]) -> None:
         pool_id = str(ioctx.get_pool_id())