osd: pglog: clean up divergent_priors off disk when running; don't assert on startup

Fixes: http://tracker.ceph.com/issues/17916
Signed-off-by: Greg Farnum <gfarnum@redhat.com>

diff --git a/src/osd/PGLog.cc b/src/osd/PGLog.cc
index e8f2a3ac49ef36e358d4bb6018ce70276b9a8055..bdb8c80b1fcd6b2620c4a2951ffecd0e7c215e79 100644 (file)
--- a/src/osd/PGLog.cc
+++ b/src/osd/PGLog.cc
@@ -177,6 +177,9 @@ void PGLog::trim(
     /* If we are trimming, we must be complete up to trim_to, time
      * to throw out any divergent_priors
      */
+    if (!divergent_priors.empty()) {
+      dirty_divergent_priors = true;
+    }
     divergent_priors.clear();
     // We shouldn't be trimming the log past last_complete
     assert(trim_to <= info.last_complete);
@@ -1044,7 +1047,20 @@ void PGLog::read_log(ObjectStore *store, coll_t pg_coll,
         * version would not have been recovered, and a newer version
         * would show up in the log above.
         */
-       assert(oi.version == i->first);
+       /**
+        * Unfortunately the assessment above is incorrect because of
+        * http://tracker.ceph.com/issues/17916 (we were incorrectly
+        * not removing the divergent_priors set from disk state!),
+        * so let's check that.
+        */
+       if (oi.version > i->first) {
+         ldpp_dout(dpp, 0) << "read_log divergent_priors entry (" << *i
+                           << ") inconsistent with disk state (" <<  oi
+                           << "), assuming it is tracker.ceph.com/issues/17916"
+                           << dendl;
+       } else {
+         assert(oi.version == i->first);
+       }
       } else {
        ldpp_dout(dpp, 15) << "read_log  missing " << *i << dendl;
        missing.add(i->second, i->first, eversion_t());