From: Sage Weil <sage@redhat.com>
Date: Wed, 10 Jul 2019 16:04:49 +0000 (-0500)
Subject: qa/tasks/ceph_manager: make is_{clean,recovered,active_or_down} less racy
X-Git-Tag: v15.1.0~2234^2
X-Git-Url: http://git.apps.os.sepia.ceph.com/?a=commitdiff_plain;h=0b4ce2ab4c19e4c2233df13bdcdb1152a6dc24ea;p=ceph-ci.git

qa/tasks/ceph_manager: make is_{clean,recovered,active_or_down} less racy

Currently these can be thrown off if the cluster is creating or removing
pools at the same time.  Fix by taking a single snapshot of the pg stats
and based our judgement on that.

Signed-off-by: Sage Weil <sage@redhat.com>
---

diff --git a/qa/tasks/ceph_manager.py b/qa/tasks/ceph_manager.py
index a127c2b3fe0..25858bf8c02 100644
--- a/qa/tasks/ceph_manager.py
+++ b/qa/tasks/ceph_manager.py
@@ -2147,6 +2147,9 @@ class CephManager:
         Find the number of active and clean pgs.
         """
         pgs = self.get_pg_stats()
+        return self._get_num_active_clean(pgs)
+
+    def _get_num_active_clean(self, pgs):
         num = 0
         for pg in pgs:
             if (pg['state'].count('active') and
@@ -2160,6 +2163,9 @@ class CephManager:
         Find the number of active and recovered pgs.
         """
         pgs = self.get_pg_stats()
+        return self._get_num_active_recovered(pgs)
+
+    def _get_num_active_recovered(self, pgs):
         num = 0
         for pg in pgs:
             if (pg['state'].count('active') and
@@ -2210,6 +2216,9 @@ class CephManager:
         Find the number of pgs that are either active or down.
         """
         pgs = self.get_pg_stats()
+        return self._get_num_active_down(pgs)
+
+    def _get_num_active_down(self, pgs):
         num = 0
         for pg in pgs:
             if ((pg['state'].count('active') and not
@@ -2236,19 +2245,22 @@ class CephManager:
         """
         True if all pgs are clean
         """
-        return self.get_num_active_clean() == self.get_num_pgs()
+        pgs = self.get_pg_stats()
+        return self._get_num_active_clean(pgs) == len(pgs)
 
     def is_recovered(self):
         """
         True if all pgs have recovered
         """
-        return self.get_num_active_recovered() == self.get_num_pgs()
+        pgs = self.get_pg_stats()
+        return self._get_num_active_recovered(pgs) == len(pgs)
 
     def is_active_or_down(self):
         """
         True if all pgs are active or down
         """
-        return self.get_num_active_down() == self.get_num_pgs()
+        pgs = self.get_pg_stats()
+        return self._get_num_active_down(pgs) == len(pgs)
 
     def wait_for_clean(self, timeout=1200):
         """