From bf4168245dec531cb75086fc42f5cbfeb80ce365 Mon Sep 17 00:00:00 2001
From: Patrick Donnelly <pdonnell@redhat.com>
Date: Thu, 16 Dec 2021 11:14:08 -0500
Subject: [PATCH] qa: test that scrub causes reintegration

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
---
 qa/tasks/cephfs/filesystem.py  |  7 ++++-
 qa/tasks/cephfs/test_strays.py | 51 ++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/qa/tasks/cephfs/filesystem.py b/qa/tasks/cephfs/filesystem.py
index 3f9b518a5a1..5dd18825f46 100644
--- a/qa/tasks/cephfs/filesystem.py
+++ b/qa/tasks/cephfs/filesystem.py
@@ -1163,7 +1163,12 @@ class Filesystem(MDSCluster):
         return self.json_asok(command, 'mds', info['name'], timeout=timeout)
 
     def rank_tell(self, command, rank=0, status=None):
-        return json.loads(self.mon_manager.raw_cluster_cmd("tell", f"mds.{self.id}:{rank}", *command))
+        try:
+            out = self.mon_manager.raw_cluster_cmd("tell", f"mds.{self.id}:{rank}", *command)
+            return json.loads(out)
+        except json.decoder.JSONDecodeError:
+            log.error("could not decode: {}".format(out))
+            raise
 
     def ranks_tell(self, command, status=None):
         if status is None:
diff --git a/qa/tasks/cephfs/test_strays.py b/qa/tasks/cephfs/test_strays.py
index 6e96ee4d529..2cf80e4de3a 100644
--- a/qa/tasks/cephfs/test_strays.py
+++ b/qa/tasks/cephfs/test_strays.py
@@ -456,6 +456,57 @@ class TestStrays(CephFSTestCase):
         # We purged it at the last
         self.assertEqual(self.get_mdc_stat("strays_enqueued"), 1)
 
+    def test_reintegration_via_scrub(self):
+        """
+        That reintegration is triggered via recursive scrub.
+        """
+
+        self.mount_a.run_shell_payload("""
+        mkdir -p a b
+        for i in `seq 1 50`; do
+           touch a/"$i"
+           ln a/"$i" b/"$i"
+        done
+        sync -f .
+        """)
+
+        self.mount_a.remount() # drop caps/cache
+        self.fs.rank_tell(["flush", "journal"])
+        self.fs.rank_fail()
+        self.fs.wait_for_daemons()
+
+        # only / in cache, reintegration cannot happen
+        self.wait_until_equal(
+            lambda: len(self.fs.rank_tell(["dump", "tree", "/"])),
+            expect_val=3,
+            timeout=60
+        )
+
+        last_reintegrated = self.get_mdc_stat("strays_reintegrated")
+        self.mount_a.run_shell_payload("""
+        rm a/*
+        sync -f .
+        """)
+        self.wait_until_equal(
+            lambda: len(self.fs.rank_tell(["dump", "tree", "/"])),
+            expect_val=3,
+            timeout=60
+        )
+        self.assertEqual(self.get_mdc_stat("num_strays"), 50)
+        curr_reintegrated = self.get_mdc_stat("strays_reintegrated")
+        self.assertEqual(last_reintegrated, curr_reintegrated)
+
+        self.fs.rank_tell(["scrub", "start", "/", "recursive,force"])
+
+        self.wait_until_equal(
+            lambda: self.get_mdc_stat("num_strays"),
+            expect_val=0,
+            timeout=60
+        )
+        curr_reintegrated = self.get_mdc_stat("strays_reintegrated")
+        # N.B.: reintegrate (rename RPC) may be tried multiple times from different code paths
+        self.assertGreaterEqual(curr_reintegrated, last_reintegrated+50)
+
     def test_mv_hardlink_cleanup(self):
         """
         That when doing a rename from A to B, and B has hardlinks,