osd: copy the reqids even if the object is deleted during promotion

If the object is deleted on the base tier, and the reqids are not copied
during promotion, this again leads to the 'ops not idempotent' problem.
For the copy-get op, this fix copies the reqids even if the object doesn't
exist.

Signed-off-by: Zhiqiang Wang <zhiqiang.wang@intel.com>

diff --git a/src/osd/ReplicatedPG.cc b/src/osd/ReplicatedPG.cc
index 224454a8ab4b2546b5c51be34308e046589a9218..329cc67dfc84e077ec1081d0c5b24875f7719f77 100644 (file)
--- a/src/osd/ReplicatedPG.cc
+++ b/src/osd/ReplicatedPG.cc
@@ -1567,6 +1567,17 @@ void ReplicatedPG::do_op(OpRequestRef& op)
     return;
 
   if (r && (r != -ENOENT || !obc)) {
+    // copy the reqids for copy get on ENOENT
+    if (r == -ENOENT &&
+       (m->ops[0].op.op == CEPH_OSD_OP_COPY_GET_CLASSIC ||
+        m->ops[0].op.op == CEPH_OSD_OP_COPY_GET)) {
+      bool classic = false;
+      if (m->ops[0].op.op == CEPH_OSD_OP_COPY_GET_CLASSIC) {
+       classic = true;
+      }
+      fill_in_copy_get_noent(op, oid, m->ops[0], classic);
+      return;
+    }
     dout(20) << __func__ << "find_object_context got error " << r << dendl;
     osd->reply_op_error(op, r);
     return;
@@ -1760,6 +1771,17 @@ void ReplicatedPG::do_op(OpRequestRef& op)
       (!obc->obs.exists ||
        ((m->get_snapid() != CEPH_SNAPDIR) &&
        obc->obs.oi.is_whiteout()))) {
+    // copy the reqids for copy get on ENOENT
+    if (m->ops[0].op.op == CEPH_OSD_OP_COPY_GET_CLASSIC ||
+       m->ops[0].op.op == CEPH_OSD_OP_COPY_GET) {
+      bool classic = false;
+      if (m->ops[0].op.op == CEPH_OSD_OP_COPY_GET_CLASSIC) {
+       classic = true;
+      }
+      fill_in_copy_get_noent(op, oid, m->ops[0], classic);
+      close_op_ctx(ctx, -ENOENT);
+      return;
+    }
     reply_ctx(ctx, -ENOENT);
     return;
   }
@@ -6382,6 +6404,28 @@ int ReplicatedPG::fill_in_copy_get(
   return result;
 }
 
+void ReplicatedPG::fill_in_copy_get_noent(OpRequestRef& op, hobject_t oid,
+                                          OSDOp& osd_op, bool classic)
+{
+  MOSDOp *m = static_cast<MOSDOp*>(op->get_req());
+  uint64_t features = m->get_features();
+  object_copy_data_t reply_obj;
+
+  pg_log.get_log().get_object_reqids(oid, 10, &reply_obj.reqids);
+  dout(20) << __func__ << " got reqids " << reply_obj.reqids << dendl;
+  if (classic) {
+    reply_obj.encode_classic(osd_op.outdata);
+  } else {
+    ::encode(reply_obj, osd_op.outdata, features);
+  }
+  osd_op.rval = -ENOENT;
+  MOSDOpReply *reply = new MOSDOpReply(m, 0, get_osdmap()->get_epoch(), 0, false);
+  reply->claim_op_out_data(m->ops);
+  reply->set_result(-ENOENT);
+  reply->add_flags(CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK);
+  osd->send_message_osd_client(reply, m->get_connection());
+}
+
 void ReplicatedPG::start_copy(CopyCallback *cb, ObjectContextRef obc,
                              hobject_t src, object_locator_t oloc,
                              version_t version, unsigned flags,

diff --git a/src/osd/ReplicatedPG.h b/src/osd/ReplicatedPG.h
index cc3ce84a4b1a82c2ab468c7b07c9b6fc9cbfc21c..81f5130fa86d2d5c0b7da9e85596e871c140c137 100644 (file)
--- a/src/osd/ReplicatedPG.h
+++ b/src/osd/ReplicatedPG.h
@@ -1354,6 +1354,8 @@ protected:
     OSDOp& op,
     ObjectContextRef& obc,
     bool classic);
+  void fill_in_copy_get_noent(OpRequestRef& op, hobject_t oid,
+                              OSDOp& osd_op, bool classic);
 
   /**
    * To copy an object, call start_copy.

diff --git a/src/osdc/Objecter.h b/src/osdc/Objecter.h
index 3e514e3e8e48cca0b44127201925b055f16f04b3..d3688afbe821ba1db3f02dc73779d3a9769779c5 100644 (file)
--- a/src/osdc/Objecter.h
+++ b/src/osdc/Objecter.h
@@ -654,12 +654,18 @@ struct ObjectOperation {
        out_flags(flags), out_data_digest(dd), out_omap_digest(od),
         out_reqids(oreqids), prval(r) {}
     void finish(int r) {
-      if (r < 0)
+      // reqids are copied on ENOENT
+      if (r < 0 && r != -ENOENT)
        return;
       try {
        bufferlist::iterator p = bl.begin();
        object_copy_data_t copy_reply;
        ::decode(copy_reply, p);
+       if (r == -ENOENT) {
+         if (out_reqids)
+           *out_reqids = copy_reply.reqids;
+         return;
+       }
        if (out_size)
          *out_size = copy_reply.size;
        if (out_mtime)