mds: trim null dentries proactively

Instead of leaving null dentries (e.g. left
behind from unlinks) in the cache until they
fall out of the LRU, actively push them
to the bottom of the LRU and then consume
all nulls at the bottom in trim() even if
the cache is not oversized yet.

This fixes the case where standby replay daemons
would otherwise accumulate a cache full of
null dentries resulting from unlinks, and it
makes the behaviour of active daemons more
deterministic.

Fixes: http://tracker.ceph.com/issues/16919
Signed-off-by: John Spray <john.spray@redhat.com>

diff --git a/src/mds/MDCache.cc b/src/mds/MDCache.cc
index 5953ed66963e67bd476cda24fbd65dc2a805338a..72eb7a95b76ed6c5235ba2d429bb89664f32f34b 100644 (file)
--- a/src/mds/MDCache.cc
+++ b/src/mds/MDCache.cc
@@ -6356,10 +6356,22 @@ bool MDCache::trim(int max, int count)
   bool is_standby_replay = mds->is_standby_replay();
   int unexpirable = 0;
   list<CDentry*> unexpirables;
-  // trim dentries from the LRU
-  while (lru.lru_get_size() + unexpirable > (unsigned)max) {
+
+  // trim dentries from the LRU: only enough to satisfy `max`,
+  // unless we see null dentries at the bottom of the LRU,
+  // in which case trim all those.
+  bool trimming_nulls = true;
+  while (trimming_nulls || lru.lru_get_size() + unexpirable > (unsigned)max) {
     CDentry *dn = static_cast<CDentry*>(lru.lru_expire());
-    if (!dn) break;
+    if (!dn) {
+      break;
+    }
+    if (!dn->get_linkage()->is_null()) {
+      trimming_nulls = false;
+      if (lru.lru_get_size() + unexpirable <= (unsigned)max) {
+        break;
+      }
+    }
     if ((is_standby_replay && dn->get_linkage()->inode &&
         dn->get_linkage()->inode->item_open_file.is_on_list()) ||
        trim_dentry(dn, expiremap)) {

diff --git a/src/mds/MDLog.cc b/src/mds/MDLog.cc
index bc018fabcc6e1458e00c85deee1df95a9e0c5ea4..ccb02a02a15f32b13510ab0f97d60e6030ffe16f 100644 (file)
--- a/src/mds/MDLog.cc
+++ b/src/mds/MDLog.cc
@@ -850,6 +850,7 @@ void MDLog::replay(MDSInternalContextBase *c)
   // empty?
   if (journaler->get_read_pos() == journaler->get_write_pos()) {
     dout(10) << "replay - journal empty, done." << dendl;
+    mds->mdcache->trim(-1);
     if (c) {
       c->complete(0);
     }

diff --git a/src/mds/journal.cc b/src/mds/journal.cc
index 3954b612b0817b3333d292ca71363a7512af483f..70a843be1df2e447612b6b86a75dc64c74758017 100644 (file)
--- a/src/mds/journal.cc
+++ b/src/mds/journal.cc
@@ -1276,7 +1276,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
        dn = dir->add_null_dentry(p->dn, p->dnfirst, p->dnlast);
        dn->set_version(p->dnv);
        if (p->is_dirty()) dn->_mark_dirty(logseg);
-       dout(10) << "EMetaBlob.replay added " << *dn << dendl;
+       dout(10) << "EMetaBlob.replay added (full) " << *dn << dendl;
       } else {
        dn->set_version(p->dnv);
        if (p->is_dirty()) dn->_mark_dirty(logseg);
@@ -1302,6 +1302,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
            mds->clog->warn(ss);
          }
          dir->unlink_inode(dn);
+          mds->mdcache->touch_dentry_bottom(dn);
        }
        if (unlinked.count(in))
          linked.insert(in);
@@ -1313,7 +1314,9 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
        if (dn->get_linkage()->get_inode() != in && in->get_parent_dn()) {
          dout(10) << "EMetaBlob.replay unlinking " << *in << dendl;
          unlinked[in] = in->get_parent_dir();
+          CDentry *unlinked_dn = in->get_parent_dn();
          in->get_parent_dir()->unlink_inode(in->get_parent_dn());
+          mds->mdcache->touch_dentry_bottom(unlinked_dn);
        }
        if (dn->get_linkage()->get_inode() != in) {
          if (!dn->get_linkage()->is_null()) { // note: might be remote.  as with stray reintegration.
@@ -1326,6 +1329,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
              mds->clog->warn(ss);
            }
            dir->unlink_inode(dn);
+            mds->mdcache->touch_dentry_bottom(dn);
          }
          if (unlinked.count(in))
            linked.insert(in);
@@ -1371,6 +1375,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
            dout(0) << ss.str() << dendl;
          }
          dir->unlink_inode(dn);
+          mds->mdcache->touch_dentry_bottom(dn);
        }
        dir->link_remote_inode(dn, p->ino, p->d_type);
        dn->set_version(p->dnv);
@@ -1392,7 +1397,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
        dn = dir->add_null_dentry(p->dn, p->dnfirst, p->dnlast);
        dn->set_version(p->dnv);
        if (p->dirty) dn->_mark_dirty(logseg);
-       dout(10) << "EMetaBlob.replay added " << *dn << dendl;
+       dout(10) << "EMetaBlob.replay added (nullbit) " << *dn << dendl;
       } else {
        dn->first = p->dnfirst;
        if (!dn->get_linkage()->is_null()) {
@@ -1405,6 +1410,7 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
            if (dn->get_linkage()->is_primary())
              unlinked[in] = dir;
            dir->unlink_inode(dn);
+            mds->mdcache->touch_dentry_bottom(dn);
          }
        }
        dn->set_version(p->dnv);
@@ -1415,6 +1421,10 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
       olddir = dir;
       if (lump.is_importing())
        dn->state_set(CDentry::STATE_AUTH);
+
+      // Make null dentries the first things we trim
+      dout(10) << "EMetaBlob.replay pushing to bottom of lru " << *dn << dendl;
+      mds->mdcache->touch_dentry_bottom(dn);
     }
   }
 
@@ -1622,7 +1632,13 @@ void EMetaBlob::replay(MDSRank *mds, LogSegment *logseg, MDSlaveUpdate *slaveup)
     CInode *in = mds->mdcache->get_inode(*p);
     if (in) {
       dout(10) << "EMetaBlob.replay destroyed " << *p << ", dropping " << *in << dendl;
+      CDentry *parent = in->get_parent_dn();
       mds->mdcache->remove_inode(in);
+      if (parent) {
+        dout(10) << "EMetaBlob.replay unlinked from dentry " << *parent << dendl;
+        assert(parent->get_linkage()->is_null());
+        mds->mdcache->touch_dentry_bottom(parent);
+      }
     } else {
       dout(10) << "EMetaBlob.replay destroyed " << *p << ", not in cache" << dendl;
     }