client: fix fuse client hang because its pipe to mds is not ok

If fuse client session had been killed by mds and the mds daemon restart
or hot-standby switch happens right away but the client did not receive
any message from monitor due to network or other whatever reason untill
the mds become active again.Thus cause client didn't do closed_mds_session
lead the seession still is STATE_OPEN but client can't send any message to
mds because its pipe is not ok.So we should close the stale session so that
it can be reopened again.

Fixes: http://tracker.ceph.com/issues/36079
Signed-off-by: Guan yunfei <yunfei.guan@xtaotech.com>
(cherry picked from commit 0e137de26e85942f8b40f7b13e564bd4c31b37f9)

Conflicts:
	src/client/Client.cc : Resolved in handle_mds_map

diff --git a/src/client/Client.cc b/src/client/Client.cc
index 27e47cce3a63eebc44f735874e13cecdee6e588f..551d06e0cb9c1c7868f20e5ddc60f31b635dde2e 100644 (file)
--- a/src/client/Client.cc
+++ b/src/client/Client.cc
@@ -2643,13 +2643,14 @@ void Client::handle_fs_map_user(MFSMapUser *m)
 
 void Client::handle_mds_map(MMDSMap* m)
 {
+  mds_gid_t old_inc, new_inc;
   if (m->get_epoch() <= mdsmap->get_epoch()) {
     ldout(cct, 1) << __func__ << " epoch " << m->get_epoch()
                   << " is identical to or older than our "
                   << mdsmap->get_epoch() << dendl;
     m->put();
     return;
-  }  
+  }
 
   ldout(cct, 1) << __func__ << " epoch " << m->get_epoch() << dendl;
 
@@ -2695,6 +2696,13 @@ void Client::handle_mds_map(MMDSMap* m)
     if (!mdsmap->is_up(mds)) {
       session->con->mark_down();
     } else if (mdsmap->get_inst(mds) != session->inst) {
+      old_inc = oldmap->get_incarnation(mds);
+      new_inc = mdsmap->get_incarnation(mds);
+      if (old_inc != new_inc) {
+        ldout(cct, 1) << "mds incarnation changed from "
+                     << old_inc << " to " << new_inc << dendl;
+        oldstate = MDSMap::STATE_NULL;
+      }
       session->con->mark_down();
       session->inst = mdsmap->get_inst(mds);
       // When new MDS starts to take over, notify kernel to trim unused entries
@@ -2705,6 +2713,11 @@ void Client::handle_mds_map(MMDSMap* m)
       continue;  // no change
     
     session->mds_state = newstate;
+    if (old_inc != new_inc && newstate > MDSMap::STATE_RECONNECT) {
+      // missed reconnect close the session so that it can be reopened
+      _closed_mds_session(session);
+      continue;
+    }
     if (newstate == MDSMap::STATE_RECONNECT) {
       session->con = messenger->get_connection(session->inst);
       send_reconnect(session);

diff --git a/src/mds/MDSMap.h b/src/mds/MDSMap.h
index 626284d1cbe0894d2eb2ec6b3e7c5b9a76512a29..566b9cfacb47c5062d041bc1338bfab5fb780d54 100644 (file)
--- a/src/mds/MDSMap.h
+++ b/src/mds/MDSMap.h
@@ -654,6 +654,16 @@ public:
     }
   }
 
+  /**
+   * Get MDS rank incarnation if the rank is up, else -1
+   */
+  mds_gid_t get_incarnation(mds_rank_t m) const {
+    std::map<mds_rank_t, mds_gid_t>::const_iterator u = up.find(m);
+    if (u == up.end())
+      return MDS_GID_NONE;
+    return (mds_gid_t)get_inc_gid(u->second);
+  }
+
   int get_inc_gid(mds_gid_t gid) const {
     auto mds_info_entry = mds_info.find(gid);
     if (mds_info_entry != mds_info.end())