osd/: bounce writes on objects written since mlcod back to primary

If there are any writes since mlcod on an object with the same head,
our ondisk state may be unstable or may reflect uncommitted state.
Bounce back to primary.

Signed-off-by: Samuel Just <sjust@redhat.com>

diff --git a/src/osd/PGLog.h b/src/osd/PGLog.h
index f39dc979c4017335c5a640e53472c89e681f7fa1..2990366f5df88fd2b488347fd9248ec4c13ce36f 100644 (file)
--- a/src/osd/PGLog.h
+++ b/src/osd/PGLog.h
@@ -329,6 +329,16 @@ public:
       return false;
     }
 
+    bool has_write_since(const hobject_t &oid, const eversion_t &bound) const {
+      for (auto i = log.rbegin(); i != log.rend(); ++i) {
+       if (i->version <= bound)
+         return false;
+       if (i->soid.get_head() == oid.get_head())
+         return true;
+      }
+      return false;
+    }
+
     /// get a (bounded) list of recent reqids for the given object
     void get_object_reqids(const hobject_t& oid, unsigned max,
                           mempool::osd_pglog::vector<pair<osd_reqid_t, version_t> > *pls,

diff --git a/src/osd/PeeringState.h b/src/osd/PeeringState.h
index c67a76881fdc416dd06d7052f2bb225920c62141..fc2719b02aee8652e3365e537dde28130d020495 100644 (file)
--- a/src/osd/PeeringState.h
+++ b/src/osd/PeeringState.h
@@ -2199,6 +2199,15 @@ public:
   bool needs_recovery() const;
   bool needs_backfill() const;
 
+  /**
+   * Returns whether a particular object can be safely read on this replica
+   */
+  bool can_serve_replica_read(const hobject_t &hoid) {
+    ceph_assert(!is_primary());
+    return !pg_log.get_log().has_write_since(
+      hoid, get_min_last_complete_ondisk());
+  }
+
   /**
    * Returns whether all peers which might have unfound objects have been
    * queried or marked lost.

diff --git a/src/osd/PrimaryLogPG.cc b/src/osd/PrimaryLogPG.cc
index 579b023914b24ed1a86cec4c35e9355f715bc59f..6e77d927851c51edc1a55e3fd9c89d24e5629aa5 100644 (file)
--- a/src/osd/PrimaryLogPG.cc
+++ b/src/osd/PrimaryLogPG.cc
@@ -2112,6 +2112,19 @@ void PrimaryLogPG::do_op(OpRequestRef& op)
     return;
   }
 
+  if (!is_primary()) {
+    if (!recovery_state.can_serve_replica_read(oid)) {
+      dout(20) << __func__ << ": oid " << oid
+              << " unstable write on replica, bouncing to primary."
+              << *m << dendl;
+      osd->reply_op_error(op, -EAGAIN);
+      return;
+    } else {
+      dout(20) << __func__ << ": serving replica read on oid" << oid
+              << dendl;
+    }
+  }
+
   int r = find_object_context(
     oid, &obc, can_create,
     m->has_flag(CEPH_OSD_FLAG_MAP_SNAP_CLONE),