mon: handle MOSDPGCreated messages

MOSDPGCreated messages are used to prune the creating_pgs_by_osd_epoch
and creating_pgs, by updating created_pools. as once a pool is created
we will not send MOSDPGCreate to its acting_primary OSD anymore.

Signed-off-by: Kefu Chai <kchai@redhat.com>

diff --git a/src/mon/Monitor.cc b/src/mon/Monitor.cc
index 19657f85d0a724152eaefebd60778c7f21f589aa..24ac15d4bcf85f7cbd37c8780af66cea4780cd49 100644 (file)
--- a/src/mon/Monitor.cc
+++ b/src/mon/Monitor.cc
@@ -3748,6 +3748,7 @@ void Monitor::dispatch_op(MonOpRequestRef op)
     case MSG_OSD_BOOT:
     case MSG_OSD_ALIVE:
     case MSG_OSD_PGTEMP:
+    case MSG_OSD_PG_CREATED:
     case MSG_REMOVE_SNAPS:
       paxos_service[PAXOS_OSDMAP]->dispatch(op);
       break;

diff --git a/src/mon/OSDMonitor.cc b/src/mon/OSDMonitor.cc
index 020c88ac15fa40edf8cfad38d36b39b00c557122..fea6682a3ce4e24a85673ab6d4a88f95682f789c 100644 (file)
--- a/src/mon/OSDMonitor.cc
+++ b/src/mon/OSDMonitor.cc
@@ -42,6 +42,7 @@
 #include "messages/MPoolOp.h"
 #include "messages/MPoolOpReply.h"
 #include "messages/MOSDPGCreate.h"
+#include "messages/MOSDPGCreated.h"
 #include "messages/MOSDPGTemp.h"
 #include "messages/MMonCommand.h"
 #include "messages/MRemoveSnaps.h"
@@ -1524,6 +1525,8 @@ bool OSDMonitor::preprocess_query(MonOpRequestRef op)
     return preprocess_boot(op);
   case MSG_OSD_ALIVE:
     return preprocess_alive(op);
+  case MSG_OSD_PG_CREATED:
+    return preprocess_pg_created(op);
   case MSG_OSD_PGTEMP:
     return preprocess_pgtemp(op);
   case MSG_OSD_BEACON:
@@ -1559,6 +1562,8 @@ bool OSDMonitor::prepare_update(MonOpRequestRef op)
     return prepare_boot(op);
   case MSG_OSD_ALIVE:
     return prepare_alive(op);
+  case MSG_OSD_PG_CREATED:
+    return prepare_pg_created(op);
   case MSG_OSD_PGTEMP:
     return prepare_pgtemp(op);
   case MSG_OSD_BEACON:
@@ -2616,6 +2621,43 @@ void OSDMonitor::_reply_map(MonOpRequestRef op, epoch_t e)
   send_latest(op, e);
 }
 
+// pg_created
+bool OSDMonitor::preprocess_pg_created(MonOpRequestRef op)
+{
+  op->mark_osdmon_event(__func__);
+  auto m = static_cast<MOSDPGCreated*>(op->get_req());
+  dout(10) << __func__ << " " << *m << dendl;
+  auto session = m->get_session();
+  if (!session) {
+    dout(10) << __func__ << ": no monitor session!" << dendl;
+    return true;
+  }
+  if (!session->is_capable("osd", MON_CAP_X)) {
+    derr << __func__ << " received from entity "
+         << "with insufficient privileges " << session->caps << dendl;
+    return true;
+  }
+  // always forward the "created!" to the leader
+  return false;
+}
+
+bool OSDMonitor::prepare_pg_created(MonOpRequestRef op)
+{
+  op->mark_osdmon_event(__func__);
+  auto m = static_cast<MOSDPGCreated*>(op->get_req());
+  dout(10) << __func__ << " " << *m << dendl;
+  auto src = m->get_orig_source();
+  auto from = src.num();
+  if (!src.is_osd() ||
+      !mon->osdmon()->osdmap.is_up(from) ||
+      m->get_orig_source_inst() != mon->osdmon()->osdmap.get_inst(from)) {
+    dout(1) << __func__ << " ignoring stats from non-active osd." << dendl;
+    return false;
+  }
+  pending_created_pgs.push_back(m->pgid);
+  return true;
+}
+
 // -------------
 // pg_temp changes
 

diff --git a/src/mon/OSDMonitor.h b/src/mon/OSDMonitor.h
index ad6966021ac936bf3b4d8f61b97d10547b41437f..54886e2420287743cd4a7a7b7071b75a657f4063 100644 (file)
--- a/src/mon/OSDMonitor.h
+++ b/src/mon/OSDMonitor.h
@@ -280,6 +280,9 @@ private:
   bool preprocess_pgtemp(MonOpRequestRef op);
   bool prepare_pgtemp(MonOpRequestRef op);
 
+  bool preprocess_pg_created(MonOpRequestRef op);
+  bool prepare_pg_created(MonOpRequestRef op);
+
   int _check_remove_pool(int64_t pool, const pg_pool_t *pi, ostream *ss);
   bool _check_become_tier(
       int64_t tier_pool_id, const pg_pool_t *tier_pool,