Merge tag 'upstream/2.0.6'

Upstream version 2.0.6

8 files changed, 86 insertions, 39 deletions

diff --git a/db/cloner.cpp b/db/cloner.cpp
index 26c2f74..6d5b095 100644
--- a/db/cloner.cpp
+++ b/db/cloner.cpp
@@ -169,7 +169,8 @@ namespace mongo {
                     getDur().commitIfNeeded();
                 }
                 catch( UserException& e ) {
-                    log() << "warning: exception cloning object in " << from_collection << ' ' << e.what() << " obj:" << js.toString() << '\n';
+                    error() << "error: exception cloning object in " << from_collection << ' ' << e.what() << " obj:" << js.toString() << '\n';
+                    throw;
                 }
 
                 RARELY if ( time( 0 ) - saveLast > 60 ) {
@@ -238,7 +239,8 @@ namespace mongo {
                     getDur().commitIfNeeded();
                 }
                 catch( UserException& e ) {
-                    log() << "warning: exception cloning object in " << from_collection << ' ' << e.what() << " obj:" << js.toString() << '\n';
+                    error() << "error: exception cloning object in " << from_collection << ' ' << e.what() << " obj:" << js.toString() << '\n';
+                    throw;
                 }
             }
         }
diff --git a/db/compact.cpp b/db/compact.cpp
index c6e5f77..c100006 100644
--- a/db/compact.cpp
+++ b/db/compact.cpp
@@ -180,6 +180,11 @@ namespace mongo {
             d->deletedList[i].writing().Null();
         }
 
+
+
+        // Start over from scratch with our extent sizing and growth
+        d->lastExtentSize=0;
+
         // before dropping indexes, at least make sure we can allocate one extent!
         uassert(14025, "compact error no space available to allocate", !allocateSpaceForANewRecord(ns, d, Record::HeaderSize+1, false).isNull());
 
diff --git a/db/dbwebserver.cpp b/db/dbwebserver.cpp
index 78c09c0..eb19ba3 100644
--- a/db/dbwebserver.cpp
+++ b/db/dbwebserver.cpp
@@ -79,11 +79,10 @@ namespace mongo {
         }
 
         bool allowed( const char * rq , vector<string>& headers, const SockAddr &from ) {
-            if ( from.isLocalHost() )
-                return true;
-
-            if ( ! _webUsers->haveAdminUsers() )
+            if ( from.isLocalHost() || !_webUsers->haveAdminUsers() ) {
+                cmdAuthenticate.authenticate( "admin", "RestUser", false );
                 return true;
+            }
 
             string auth = getHeader( rq , "Authorization" );
 
@@ -118,8 +117,10 @@ namespace mongo {
                     r << ha2;
                     string r1 = md5simpledigest( r.str() );
 
-                    if ( r1 == parms["response"] )
+                    if ( r1 == parms["response"] ) {
+                        cmdAuthenticate.authenticate( "admin", user["user"].str(), user[ "readOnly" ].isBoolean() && user[ "readOnly" ].boolean() );
                         return true;
+                    }
                 }
             }
 
diff --git a/db/queryoptimizer.cpp b/db/queryoptimizer.cpp
index 71ca657..8ec4cb4 100644
--- a/db/queryoptimizer.cpp
+++ b/db/queryoptimizer.cpp
@@ -214,22 +214,6 @@ doneCheckOrder:
         if ( willScanTable() ) {
             if ( _frs.nNontrivialRanges() ) {
                 checkTableScanAllowed( _frs.ns() );
-                
-                // if we are doing a table scan on _id
-                // and its a capped collection
-                // we disallow as its a common user error
-                // .system. and local collections are exempt
-                if ( _d && _d->capped && _frs.range( "_id" ).nontrivial() ) {
-                    if ( cc().isSyncThread() ||
-                         str::contains( _frs.ns() , ".system." ) || 
-                         str::startsWith( _frs.ns() , "local." ) ) {
-                        // ok
-                    }
-                    else {
-                        warning() << "_id query on capped collection without an _id index, performance will be poor collection: " << _frs.ns() << endl;
-                        //uassert( 14820, str::stream() << "doing _id query on a capped collection without an index is not allowed: " << _frs.ns() ,
-                    }
-                }
             }
             return findTableScan( _frs.ns(), _order, startLoc );
         }
@@ -486,12 +470,14 @@ doneCheckOrder:
                     _usingPrerecordedPlan = true;
                     _mayRecordPlan = false;
                     _plans.push_back( p );
+                    warnOnCappedIdTableScan();
                     return;
                 }
             }
         }
 
         addOtherPlans( false );
+        warnOnCappedIdTableScan();
     }
 
     void QueryPlanSet::addOtherPlans( bool checkFirst ) {
@@ -633,6 +619,31 @@ doneCheckOrder:
         }
         return _plans[0];
     }
+    
+    void QueryPlanSet::warnOnCappedIdTableScan() const {
+        // if we are doing a table scan on _id
+        // and it's a capped collection
+        // we warn as it's a common user error
+        // .system. and local collections are exempt
+        const char *ns = _frsp->ns();
+        NamespaceDetails *d = nsdetails( ns );
+        if ( d &&
+            d->capped &&
+            nPlans() == 1 &&
+            firstPlan()->willScanTable() &&
+            firstPlan()->multikeyFrs().range( "_id" ).nontrivial() ) {
+            if ( cc().isSyncThread() ||
+                str::contains( ns , ".system." ) ||
+                str::startsWith( ns , "local." ) ) {
+                // ok
+            }
+            else {
+                warning()
+                << "unindexed _id query on capped collection, "
+                << "performance will be poor collection: " << ns << endl;
+            }
+        }
+    }
 
     QueryPlanSet::Runner::Runner( QueryPlanSet &plans, QueryOp &op ) :
         _op( op ),
@@ -1247,8 +1258,12 @@ doneCheckOrder:
     void QueryUtilIndexed::clearIndexesForPatterns( const FieldRangeSetPair &frsp, const BSONObj &order ) {
         SimpleMutex::scoped_lock lk(NamespaceDetailsTransient::_qcMutex);
         NamespaceDetailsTransient& nsd = NamespaceDetailsTransient::get_inlock( frsp.ns() );
-        nsd.registerIndexForPattern( frsp._singleKey.pattern( order ), BSONObj(), 0 );
-        nsd.registerIndexForPattern( frsp._multiKey.pattern( order ), BSONObj(), 0 );
+        if ( frsp._singleKey.matchPossible() ) {
+            nsd.registerIndexForPattern( frsp._singleKey.pattern( order ), BSONObj(), 0 );
+        }
+        if ( frsp._multiKey.matchPossible() ) {
+            nsd.registerIndexForPattern( frsp._multiKey.pattern( order ), BSONObj(), 0 );
+        }
     }
     
     pair< BSONObj, long long > QueryUtilIndexed::bestIndexForPatterns( const FieldRangeSetPair &frsp, const BSONObj &order ) {
diff --git a/db/queryoptimizer.h b/db/queryoptimizer.h
index fea6c0b..78d169d 100644
--- a/db/queryoptimizer.h
+++ b/db/queryoptimizer.h
@@ -314,6 +314,7 @@ namespace mongo {
         }
         void init();
         void addHint( IndexDetails &id );
+        void warnOnCappedIdTableScan() const;
         class Runner {
         public:
             Runner( QueryPlanSet &plans, QueryOp &op );
diff --git a/db/repl.cpp b/db/repl.cpp
index 5edf0c2..8dcdd13 100644
--- a/db/repl.cpp
+++ b/db/repl.cpp
@@ -1115,7 +1115,7 @@ namespace mongo {
 
     bool OplogReader::commonConnect(const string& hostName) {
         if( conn() == 0 ) {
-            _conn = shared_ptr<DBClientConnection>(new DBClientConnection( false, 0, 0 /* tcp timeout */));
+            _conn = shared_ptr<DBClientConnection>(new DBClientConnection( false, 0, 60*10 /* tcp timeout */));
             string errmsg;
             ReplInfo r("trying to connect to sync source");
             if ( !_conn->connect(hostName.c_str(), errmsg) ||
diff --git a/db/repl/rs_initialsync.cpp b/db/repl/rs_initialsync.cpp
index 112d739..7065487 100644
--- a/db/repl/rs_initialsync.cpp
+++ b/db/repl/rs_initialsync.cpp
@@ -43,18 +43,24 @@ namespace mongo {
     }
 
     void ReplSetImpl::syncDoInitialSync() {
+        const static int maxFailedAttempts = 3;
         createOplog();
-
-        while( 1 ) {
+        int failedAttempts = 0;
+        while ( failedAttempts < maxFailedAttempts ) {
             try {
                 _syncDoInitialSync();
                 break;
             }
             catch(DBException& e) {
-                sethbmsg("initial sync exception " + e.toString(), 0);
+                failedAttempts++;
+                str::stream msg;
+                msg << "initial sync exception: ";
+                msg << e.toString() << " " << (maxFailedAttempts - failedAttempts) << " attempts remaining" ;
+                sethbmsg(msg, 0);
                 sleepsecs(30);
             }
         }
+        if ( failedAttempts >= maxFailedAttempts ) ::abort();
     }
 
     /* todo : progress metering to sethbmsg. */
@@ -80,7 +86,7 @@ namespace mongo {
     }
 
     const Member* ReplSetImpl::getMemberToSyncTo() {
-        Member *closest = 0;
+
         bool buildIndexes = true;
 
         // wait for 2N pings before choosing a sync target
@@ -95,16 +101,31 @@ namespace mongo {
             buildIndexes = myConfig().buildIndexes;
         }
 
+        Member *closest = 0;
+
         // find the member with the lowest ping time that has more data than me
-        for (Member *m = _members.head(); m; m = m->next()) {
-            if (m->hbinfo().up() &&
-                // make sure members with buildIndexes sync from other members w/indexes
-                (!buildIndexes || (buildIndexes && m->config().buildIndexes)) &&
-                (m->state() == MemberState::RS_PRIMARY ||
-                 (m->state() == MemberState::RS_SECONDARY && m->hbinfo().opTime > lastOpTimeWritten)) &&
-                (!closest || m->hbinfo().ping < closest->hbinfo().ping)) {
-                closest = m;
+
+        // Make two attempts.  The first attempt, we ignore those nodes with
+        // slave delay higher than our own.  The second attempt includes such
+        // nodes, in case those are the only ones we can reach.
+        for (int attempts = 0; attempts < 2; ++attempts) {
+            for (Member *m = _members.head(); m; m = m->next()) {
+                if (m->hbinfo().up() &&
+                    // make sure members with buildIndexes sync from other members w/indexes
+                    (!buildIndexes || (buildIndexes && m->config().buildIndexes)) &&
+                    (m->state() == MemberState::RS_PRIMARY ||
+                     (m->state() == MemberState::RS_SECONDARY && 
+                      m->hbinfo().opTime > lastOpTimeWritten)) &&
+                    (!closest || m->hbinfo().ping < closest->hbinfo().ping)) {
+
+                    if ( attempts == 0 && 
+                         myConfig().slaveDelay < m->config().slaveDelay ) {
+                        break; // skip this one in the first attempt
+                    }
+                    closest = m;
+                }
             }
+            if (closest) break; // no need for second attempt
         }
 
         {
diff --git a/db/security_common.h b/db/security_common.h
index c9a3e3a..80a7450 100644
--- a/db/security_common.h
+++ b/db/security_common.h
@@ -61,10 +61,12 @@ namespace mongo {
         virtual void help(stringstream& ss) const { ss << "internal"; }
         CmdAuthenticate() : Command("authenticate") {}
         bool run(const string& dbname , BSONObj& cmdObj, int options, string& errmsg, BSONObjBuilder& result, bool fromRepl);
+        void authenticate(const string& dbname, const string& user, const bool readOnly);
     private:
         bool getUserObj(const string& dbname, const string& user, BSONObj& userObj, string& pwd);
-        void authenticate(const string& dbname, const string& user, const bool readOnly);
     };
+    
+    extern CmdAuthenticate cmdAuthenticate;
 
     class CmdLogout : public Command {
     public: