osd: add object to missing if we find it missing on disk

If the recovery finds the object missing on disk during recovery, add it
to the local missing set so we can (hopefully) recover it from another
replica.

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

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index ee00a0099ee316753584e9d8afe30ae5e5a32a66..ab9b400e4da5b3233ef960633a22bb74a6584280 100644 (file)
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -137,7 +137,16 @@ void ReplicatedPG::wait_for_degraded_object(const sobject_t& soid, Message *m)
            << soid 
            << ", pushing"
            << dendl;
-    recover_object_replicas(soid);
+    eversion_t v;
+    for (unsigned i = 1; i < acting.size(); i++) {
+      int peer = acting[i];
+      if (peer_missing.count(peer) &&
+         peer_missing[peer].missing.count(soid)) {
+       v = peer_missing[peer].missing[soid].need;
+       break;
+      }
+    }
+    recover_object_replicas(soid, v);
   }
   waiting_for_degraded_object[soid].push_back(m);
 }
@@ -3453,7 +3462,7 @@ int ReplicatedPG::send_push_op(const sobject_t& soid, eversion_t version, int pe
   object_info_t oi(bv);
 
   if (oi.version != version) {
-    osd->clog.error() << "push " << soid << " v " << version << " to osd" << peer
+    osd->clog.error() << info.pgid << " push " << soid << " v " << version << " to osd" << peer
                      << " failed because local copy is " << oi.version << "\n";
     return -1;
   }
@@ -3578,9 +3587,8 @@ void ReplicatedPG::sub_op_pull(MOSDSubOp *op)
   struct stat st;
   int r = osd->store->stat(coll, soid, &st);
   if (r != 0) {
-    osd->clog.error() << op->get_source() << " tried to pull " << soid
-       << " in " << info.pgid << " but got "
-       << cpp_strerror(-r) << "\n";
+    osd->clog.error() << info.pgid << " " << op->get_source() << " tried to pull " << soid
+                     << " but got " << cpp_strerror(-r) << "\n";
     send_push_op_blank(soid, op->get_source().num());
   } else {
     uint64_t size = st.st_size;
@@ -4228,7 +4236,7 @@ int ReplicatedPG::recover_primary(int max)
   return started;
 }
 
-int ReplicatedPG::recover_object_replicas(const sobject_t& soid)
+int ReplicatedPG::recover_object_replicas(const sobject_t& soid, eversion_t v)
 {
   int started = 0;
 
@@ -4237,7 +4245,16 @@ int ReplicatedPG::recover_object_replicas(const sobject_t& soid)
   // NOTE: we know we will get a valid oloc off of disk here.
   ObjectContext *obc = get_object_context(soid, OLOC_BLANK, false);
   if (!obc) {
-    osd->clog.error() << "missing primary copy of " << soid << "\n";
+    osd->clog.error() << info.pgid << " missing primary copy of " << soid << "\n";
+    missing.add(soid, v, eversion_t());
+    for (unsigned i=1; i<acting.size(); i++) {
+      int peer = acting[i];
+      if (!peer_missing[peer].is_missing(soid, v)) {
+       dout(10) << info.pgid << " unexpectedly missing " << soid << " v" << v
+                << ", will use copy on osd" << peer << dendl;
+       missing_loc[soid].insert(peer);
+      }
+    }
     return 0;
   }
 
@@ -4284,6 +4301,7 @@ int ReplicatedPG::recover_replicas(int max)
           p != m.rmissing.end() && started < max;
           ++p) {
       const sobject_t soid(p->second);
+      eversion_t v = p->first;
       if (pushing.count(soid)) {
        dout(10) << __func__ << ": already pushing " << soid << dendl;
        continue;
@@ -4297,7 +4315,7 @@ int ReplicatedPG::recover_replicas(int max)
       }
 
       dout(10) << __func__ << ": recover_object_replicas(" << soid << ")" << dendl;
-      started += recover_object_replicas(soid);
+      started += recover_object_replicas(soid, v);
     }
   }
 

diff --git a/src/osd/ReplicatedPG.h b/src/osd/ReplicatedPG.h
index ab73f7e99511394a0729f1a189309943185495ea..45147cbcc585c012f21449f2e45036aed8f4be7b 100644 (file)
--- a/src/osd/ReplicatedPG.h
+++ b/src/osd/ReplicatedPG.h
@@ -511,7 +511,7 @@ protected:
   };
   map<sobject_t, map<int, push_info_t> > pushing;
 
-  int recover_object_replicas(const sobject_t& soid);
+  int recover_object_replicas(const sobject_t& soid, eversion_t v);
   void calc_head_subsets(SnapSet& snapset, const sobject_t& head,
                         Missing& missing,
                         interval_set<uint64_t>& data_subset,