mds: handle interaction between slave rollback and fragmenting

For slave rename and rmdir events, the MDS needs to preserve non-auth
dirfrag where the renamed inode originally lives in until slave commit
event is encountered. Current method to handle this is use MDCache::
uncommitted_slave_rename_olddir to track any non-auth dirfrag that
need to be preserved. This method does not works well if any preserved
dirfrag gets fragmented by log event (such as ESubtreeMap) between the
slave prepare event and the slave commit event.

The fix is tracking inode of dirfrag instead of tracking dirfrag that
need to preserved directly.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>

diff --git a/src/mds/MDCache.cc b/src/mds/MDCache.cc
index a26633a44d216fcc07bbc7eb0119f80f8b88e823..92361ff296284028fbc65e2daa2e7240428bcd55 100644 (file)
--- a/src/mds/MDCache.cc
+++ b/src/mds/MDCache.cc
@@ -3163,7 +3163,7 @@ void MDCache::add_uncommitted_slave_update(metareqid_t reqid, int master, MDSlav
 {
   assert(uncommitted_slave_updates[master].count(reqid) == 0);
   uncommitted_slave_updates[master][reqid] = su;
-  for(set<CDir*>::iterator p = su->olddirs.begin(); p != su->olddirs.end(); ++p)
+  for(set<CInode*>::iterator p = su->olddirs.begin(); p != su->olddirs.end(); ++p)
     uncommitted_slave_rename_olddir[*p]++;
   for(set<CInode*>::iterator p = su->unlinked.begin(); p != su->unlinked.end(); ++p)
     uncommitted_slave_unlink[*p]++;
@@ -3178,25 +3178,38 @@ void MDCache::finish_uncommitted_slave_update(metareqid_t reqid, int master)
   if (uncommitted_slave_updates[master].empty())
     uncommitted_slave_updates.erase(master);
   // discard the non-auth subtree we renamed out of
-  for(set<CDir*>::iterator p = su->olddirs.begin(); p != su->olddirs.end(); ++p) {
-    CDir *dir = *p;
-    uncommitted_slave_rename_olddir[dir]--;
-    if (uncommitted_slave_rename_olddir[dir] == 0) {
-      uncommitted_slave_rename_olddir.erase(dir);
-      CDir *root = get_subtree_root(dir);
-      if (root->get_dir_auth() == CDIR_AUTH_UNDEF)
-       try_trim_non_auth_subtree(root);
-    }
+  for(set<CInode*>::iterator p = su->olddirs.begin(); p != su->olddirs.end(); ++p) {
+    CInode *diri = *p;
+    map<CInode*, int>::iterator it = uncommitted_slave_rename_olddir.find(diri);
+    assert(it != uncommitted_slave_rename_olddir.end());
+    it->second--;
+    if (it->second == 0) {
+      uncommitted_slave_rename_olddir.erase(it);
+      list<CDir*> ls;
+      diri->get_dirfrags(ls);
+      for (list<CDir*>::iterator q = ls.begin(); q != ls.end(); ++q) {
+       CDir *root = get_subtree_root(*q);
+       if (root->get_dir_auth() == CDIR_AUTH_UNDEF) {
+         try_trim_non_auth_subtree(root);
+         if (*q != root)
+           break;
+       }
+      }
+    } else
+      assert(it->second > 0);
   }
   // removed the inodes that were unlinked by slave update
   for(set<CInode*>::iterator p = su->unlinked.begin(); p != su->unlinked.end(); ++p) {
     CInode *in = *p;
-    uncommitted_slave_unlink[in]--;
-    if (uncommitted_slave_unlink[in] == 0) {
-      uncommitted_slave_unlink.erase(in);
+    map<CInode*, int>::iterator it = uncommitted_slave_unlink.find(in);
+    assert(it != uncommitted_slave_unlink.end());
+    it->second--;
+    if (it->second == 0) {
+      uncommitted_slave_unlink.erase(it);
       if (!in->get_projected_parent_dn())
        mds->mdcache->remove_inode_recursive(in);
-    }
+    } else
+      assert(it->second > 0);
   }
   delete su;
 }
@@ -6673,7 +6686,7 @@ bool MDCache::trim_non_auth_subtree(CDir *dir)
 {
   dout(10) << "trim_non_auth_subtree(" << dir << ") " << *dir << dendl;
 
-  if (uncommitted_slave_rename_olddir.count(dir) || // preserve the dir for rollback
+  if (uncommitted_slave_rename_olddir.count(dir->inode) || // preserve the dir for rollback
       my_ambiguous_imports.count(dir->dirfrag()))
     return true;
 
@@ -6694,7 +6707,7 @@ bool MDCache::trim_non_auth_subtree(CDir *dir)
         for (list<CDir*>::iterator subdir = subdirs.begin();
             subdir != subdirs.end();
             ++subdir) {
-          if (uncommitted_slave_rename_olddir.count(*subdir) || // preserve the dir for rollback
+          if (uncommitted_slave_rename_olddir.count((*subdir)->inode) || // preserve the dir for rollback
              my_ambiguous_imports.count((*subdir)->dirfrag()) ||
              (*subdir)->is_subtree_root()) {
             keep_inode = true;

diff --git a/src/mds/MDCache.h b/src/mds/MDCache.h
index 49c609511a98c12726fbddd52520f3481d78459d..d6b81795604ce6dd0b31c019d60e4dbd54f0a3aa 100644 (file)
--- a/src/mds/MDCache.h
+++ b/src/mds/MDCache.h
@@ -317,7 +317,7 @@ protected:
   map<int, map<dirfrag_t, vector<dirfrag_t> > > other_ambiguous_imports;  
 
   map<int, map<metareqid_t, MDSlaveUpdate*> > uncommitted_slave_updates;  // slave: for replay.
-  map<CDir*, int> uncommitted_slave_rename_olddir;  // slave: preserve the non-auth dir until seeing commit.
+  map<CInode*, int> uncommitted_slave_rename_olddir;  // slave: preserve the non-auth dir until seeing commit.
   map<CInode*, int> uncommitted_slave_unlink;  // slave: preserve the unlinked inode until seeing commit.
 
   // track master requests whose slaves haven't acknowledged commit
@@ -617,6 +617,13 @@ public:
       return NULL;
     return in->get_dirfrag(df.frag);
   }
+  CDir* get_dirfrag(inodeno_t ino, const string& dn) {
+    CInode *in = get_inode(ino);
+    if (!in)
+      return NULL;
+    frag_t fg = in->pick_dirfrag(dn);
+    return in->get_dirfrag(fg);
+  }
   CDir* get_force_dirfrag(dirfrag_t df) {
     CInode *diri = get_inode(df.ino);
     if (!diri)

diff --git a/src/mds/Mutation.h b/src/mds/Mutation.h
index bc60a776597dc404370ce2484eace03c885e1948..92750bef2f40ea37a4cf02cebafda0a3deccb914 100644 (file)
--- a/src/mds/Mutation.h
+++ b/src/mds/Mutation.h
@@ -328,7 +328,7 @@ struct MDSlaveUpdate {
   bufferlist rollback;
   elist<MDSlaveUpdate*>::item item;
   Context *waiter;
-  set<CDir*> olddirs;
+  set<CInode*> olddirs;
   set<CInode*> unlinked;
   MDSlaveUpdate(int oo, bufferlist &rbl, elist<MDSlaveUpdate*> &list) :
     origop(oo),

diff --git a/src/mds/Server.cc b/src/mds/Server.cc
index 7416319be0e1110cf1bca05cd009bdd632ee959b..47af825736ede6f43c992fc8b3a52bc067925fd6 100644 (file)
--- a/src/mds/Server.cc
+++ b/src/mds/Server.cc
@@ -5304,6 +5304,8 @@ void Server::do_rmdir_rollback(bufferlist &rbl, int master, MDRequest *mdr)
   assert(mdr || mds->is_resolve());
 
   CDir *dir = mds->mdcache->get_dirfrag(rollback.src_dir);
+  if (!dir)
+    dir = mds->mdcache->get_dirfrag(rollback.src_dir.ino, rollback.src_dname);
   assert(dir);
   CDentry *dn = dir->lookup(rollback.src_dname);
   assert(dn);
@@ -6928,6 +6930,8 @@ void Server::do_rename_rollback(bufferlist &rbl, int master, MDRequest *mdr,
 
   CDentry *srcdn = NULL;
   CDir *srcdir = mds->mdcache->get_dirfrag(rollback.orig_src.dirfrag);
+  if (!srcdir)
+    srcdir = mds->mdcache->get_dirfrag(rollback.orig_src.dirfrag.ino, rollback.orig_src.dname);
   if (srcdir) {
     dout(10) << "  srcdir " << *srcdir << dendl;
     srcdn = srcdir->lookup(rollback.orig_src.dname);
@@ -6941,6 +6945,8 @@ void Server::do_rename_rollback(bufferlist &rbl, int master, MDRequest *mdr,
 
   CDentry *destdn = NULL;
   CDir *destdir = mds->mdcache->get_dirfrag(rollback.orig_dest.dirfrag);
+  if (!destdir)
+    destdir = mds->mdcache->get_dirfrag(rollback.orig_dest.dirfrag.ino, rollback.orig_dest.dname);
   if (destdir) {
     dout(10) << " destdir " << *destdir << dendl;
     destdn = destdir->lookup(rollback.orig_dest.dname);

diff --git a/src/mds/journal.cc b/src/mds/journal.cc
index c7550e746758d52e18aaaeaad86c0805ef495e3c..0c6f6694a2314956a1c44b86930a9acaf6d43684 100644 (file)
--- a/src/mds/journal.cc
+++ b/src/mds/journal.cc
@@ -1208,7 +1208,7 @@ void EMetaBlob::replay(MDS *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
          assert(dir);
          // preserve subtree bound until slave commit
          if (dir->get_dir_auth() == CDIR_AUTH_UNDEF)
-           slaveup->olddirs.insert(dir);
+           slaveup->olddirs.insert(dir->inode);
        }
       }
 
@@ -1218,7 +1218,7 @@ void EMetaBlob::replay(MDS *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
       CDir *root = mds->mdcache->get_subtree_root(olddir);
       if (root->get_dir_auth() == CDIR_AUTH_UNDEF) {
        if (slaveup) // preserve the old dir until slave commit
-         slaveup->olddirs.insert(olddir);
+         slaveup->olddirs.insert(olddir->inode);
        else
          mds->mdcache->try_trim_non_auth_subtree(root);
       }