Merge branch 'master' of github.com:mongodb/mongo

7 files changed, 133 insertions, 33 deletions

diff --git a/jstests/replsets/replset1.js b/jstests/replsets/replset1.js
index 61f3d99d549..6a18dffaaa2 100644
--- a/jstests/replsets/replset1.js
+++ b/jstests/replsets/replset1.js
@@ -70,9 +70,6 @@ doTest = function( signal ) {
 
     }
 
-    // Here's how to restart a node:
-    replTest.restart( master_id );
-
     // Now let's write some documents to the new master
     for(var i=0; i<1000; i++) {
         new_master.getDB("bar").bar.save({a: i});
@@ -82,6 +79,7 @@ doTest = function( signal ) {
     // Here's how to restart the old master node:
     slave = replTest.restart( master_id );
 
+
     // Now, let's make sure that the old master comes up as a slave
     assert.soon(function() {
         var res = slave.getDB("admin").runCommand({ismaster: 1});
@@ -98,6 +96,7 @@ doTest = function( signal ) {
 
     // And that both slave nodes have all the updates
     new_master = replTest.getMaster();
+    assert.eq( 1000 , new_master.getDB( "bar" ).runCommand( { count:"bar"} ).n , "assumption 2")
     replTest.awaitReplication();
 
     slaves = replTest.liveNodes.slaves;
@@ -106,11 +105,11 @@ doTest = function( signal ) {
         slave.setSlaveOk();
         var count = slave.getDB("bar").runCommand({count: "bar"});
         printjson( count );
-        assert( count.n == 1000 , slave + " expected 1000 but count was " + count.n);
+        assert.eq( 1000 , count.n , "slave count wrong: " + slave );
     });
 
     // Shut down the set and finish the test.
     replTest.stopSet( signal );
 }
 
-// doTest( 15 );
+doTest( 15 );
diff --git a/s/config.cpp b/s/config.cpp
index 50682c67cb2..83ff5d9c545 100644
--- a/s/config.cpp
+++ b/s/config.cpp
@@ -431,11 +431,95 @@ namespace mongo {
             if ( ! ok )
                 return false;
         }
+
+        _config = configHosts;
         
         string fullString;
         joinStringDelim( configHosts, &fullString, ',' );
         _primary.setAddress( fullString , true );
         log(1) << " config string : " << fullString << endl;
+
+        return true;
+    }
+
+    bool ConfigServer::checkConfigServersConsistent( string& errmsg , int tries ) const {
+        if ( _config.size() == 1 )
+            return true;
+        
+        if ( tries <= 0 )
+            return false;
+        
+        unsigned firstGood = 0;
+        int up = 0;
+        vector<BSONObj> res;
+        for ( unsigned i=0; i<_config.size(); i++ ){
+            BSONObj x;
+            try {
+                ScopedDbConnection conn( _config[i] );
+                if ( ! conn->simpleCommand( "config" , &x , "dbhash" ) )
+                    x = BSONObj();
+                else {
+                    x = x.getOwned();
+                    if ( up == 0 )
+                        firstGood = i;
+                    up++;
+                }
+                conn.done();
+            }
+            catch ( std::exception& e ){
+                log(LL_WARNING) << " couldn't check on config server:" << _config[i] << " ok for now" << endl;
+            }
+            res.push_back(x);
+        }
+
+        if ( up == 0 ){
+            errmsg = "no config servers reachable";
+            return false;
+        }
+
+        if ( up == 1 ){
+            log( LL_WARNING ) << "only 1 config server reachable, continuing" << endl;
+            return true;
+        }
+
+        BSONObj base = res[firstGood];
+        for ( unsigned i=firstGood+1; i<res.size(); i++ ){
+            if ( res[i].isEmpty() )
+                continue;
+
+            string c1 = base.getFieldDotted( "collections.chunks" );
+            string c2 = res[i].getFieldDotted( "collections.chunks" );
+            
+            string d1 = base.getFieldDotted( "collections.databases" );
+            string d2 = res[i].getFieldDotted( "collections.databases" );
+
+            if ( c1 == c2 && d1 == d2 )
+                continue;
+            
+            stringstream ss;
+            ss << "config servers " << _config[firstGood] << " and " << _config[i] << " differ";
+            log( LL_WARNING ) << ss.str();
+            if ( tries <= 1 ){
+                ss << "\n" << c1 << "\t" << c2 << "\n" << d1 << "\t" << d2;
+                errmsg = ss.str();
+                return false;
+            }
+            
+            return checkConfigServersConsistent( errmsg , tries - 1 );
+        }
+        
+        return true;
+    }
+
+    bool ConfigServer::ok( bool checkConsistency ){
+        if ( ! _primary.ok() )
+            return false;
+        
+        string errmsg;
+        if ( ! checkConfigServersConsistent( errmsg ) ){
+            log( LL_ERROR ) << "config servers not in sync! " << errmsg << endl;
+            return false;
+        }
         
         return true;
     }
@@ -513,10 +597,15 @@ namespace mongo {
         
         
         // indexes
-        conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "min" << 1 ) , true );
-        conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "shard" << 1 << "min" << 1 ) , true );
-        conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "lastmod" << 1 ) , true );
-        conn->ensureIndex( ShardNS::shard , BSON( "host" << 1 ) , true );
+        try {
+            conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "min" << 1 ) , true );
+            conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "shard" << 1 << "min" << 1 ) , true );
+            conn->ensureIndex( ShardNS::chunk , BSON( "ns" << 1 << "lastmod" << 1 ) , true );
+            conn->ensureIndex( ShardNS::shard , BSON( "host" << 1 ) , true );
+        }
+        catch ( std::exception& e ){
+            log( LL_WARNING ) << "couldn't create indexes on config db: " << e.what() << endl;
+        }
 
         conn.done();
     }
diff --git a/s/config.h b/s/config.h
index c1767743acd..5bff03f2711 100644
--- a/s/config.h
+++ b/s/config.h
@@ -199,9 +199,7 @@ namespace mongo {
         ConfigServer();
         ~ConfigServer();
         
-        bool ok(){
-            return _primary.ok();
-        }
+        bool ok( bool checkConsistency = false );
         
         virtual string modelServer(){
             uassert( 10190 ,  "ConfigServer not setup" , _primary.ok() );
@@ -241,8 +239,16 @@ namespace mongo {
 
         static int VERSION;
         
+
+        /**
+         * check to see if all config servers have the same state
+         * will try tries time to make sure not catching in a bad state
+         */
+        bool checkConfigServersConsistent( string& errmsg , int tries = 4 ) const;
+
     private:
         string getHost( string name , bool withPort );
+        vector<string> _config;
     };
 
 } // namespace mongo
diff --git a/s/grid.cpp b/s/grid.cpp
index ea05f1bcc99..6843059c8c9 100644
--- a/s/grid.cpp
+++ b/s/grid.cpp
@@ -132,6 +132,12 @@ namespace mongo {
         try {
             ScopedDbConnection newShardConn( host );
             newShardConn->getLastError();
+            
+            if ( newShardConn->type() == ConnectionString::SYNC ){
+                newShardConn.done();
+                *errMsg = "can't use sync cluster as a shard.  for replica set, have to use <name>/<server1>,<server2>,...";
+                return false;
+            }
 
             // get the shard's local db's listing
             BSONObj res;
diff --git a/s/server.cpp b/s/server.cpp
index 09cdb0fbbfc..75c6c782bd8 100644
--- a/s/server.cpp
+++ b/s/server.cpp
@@ -260,7 +260,7 @@ int main(int argc, char* argv[], char *envp[] ) {
         return 7;
     }
     
-    if ( ! configServer.ok() ){
+    if ( ! configServer.ok( true ) ){
         cout << "configServer startup check failed" << endl;
         return 8;
     }
diff --git a/shell/servers.js b/shell/servers.js
index 51f3d8af3c6..c36fba4bba9 100644
--- a/shell/servers.js
+++ b/shell/servers.js
@@ -1199,7 +1199,7 @@ ReplSetTest.prototype.reInitiate = function() {
 ReplSetTest.prototype.awaitReplication = function() {
    this.getMaster();
 
-   latest = this.liveNodes.master.getDB("local")['oplog.rs'].find({}).sort({'$natural': -1}).limit(1).next()['ts']['t']
+   latest = this.liveNodes.master.getDB("local")['oplog.rs'].find({}).sort({'$natural': -1}).limit(1).next()['ts']
    print(latest);
 
    this.attempt({context: this, timeout: 30000, desc: "awaiting replication"},
@@ -1220,10 +1220,10 @@ ReplSetTest.prototype.awaitReplication = function() {
              if(log.find({}).sort({'$natural': -1}).limit(1).hasNext()) {
                var entry = log.find({}).sort({'$natural': -1}).limit(1).next();
                printjson( entry );
-               var ts = entry['ts']['t'];
+               var ts = entry['ts'];
                print("TS for " + slave + " is " + ts + " and latest is " + latest);
                print("Oplog size for " + slave + " is " + log.count());
-               synced = (synced && (latest == ts));
+               synced = (synced && friendlyEqual(latest,ts))
              }
              else {
                synced = false;
diff --git a/shell/utils.js b/shell/utils.js
index 65fb52ed073..de26403afa6 100644
--- a/shell/utils.js
+++ b/shell/utils.js
@@ -1046,23 +1046,23 @@ Geo.distance = function( a , b ){
                       Math.pow( bx - ax , 2 ) );
 }
 
-rs = function () { return "try rs.help()"; }

-

-rs.help = function () {

-    print("\trs.status()                     { replSetGetStatus : 1 } checks repl set status");

-    print("\trs.initiate()                   { replSetInitiate : null } initiates set with default settings");

-    print("\trs.initiate(cfg)                { replSetInitiate : cfg } initiates set with configuration cfg");

-    print("\trs.add(hostportstr)             add a new member to the set with default attributes");

-    print("\trs.add(membercfgobj)            add a new member to the set with extra attributes");

-    print("\trs.addArb(hostportstr)          add a new member which is arbiterOnly:true");

-    print("\trs.stepDown()                   step down as primary (momentarily)");

-    print("\trs.conf()                       return configuration from local.system.replset");

-    print("\trs.slaveOk()                    shorthand for db.getMongo().setSlaveOk()");

-    print();

-    print("\tdb.isMaster()                   check who is primary");

-    print();

-    print("\tsee also http://<mongod_host>:28017/_replSet for additional diagnostic info");

-}

+rs = function () { return "try rs.help()"; }
+
+rs.help = function () {
+    print("\trs.status()                     { replSetGetStatus : 1 } checks repl set status");
+    print("\trs.initiate()                   { replSetInitiate : null } initiates set with default settings");
+    print("\trs.initiate(cfg)                { replSetInitiate : cfg } initiates set with configuration cfg");
+    print("\trs.add(hostportstr)             add a new member to the set with default attributes");
+    print("\trs.add(membercfgobj)            add a new member to the set with extra attributes");
+    print("\trs.addArb(hostportstr)          add a new member which is arbiterOnly:true");
+    print("\trs.stepDown()                   step down as primary (momentarily)");
+    print("\trs.conf()                       return configuration from local.system.replset");
+    print("\trs.slaveOk()                    shorthand for db.getMongo().setSlaveOk()");
+    print();
+    print("\tdb.isMaster()                   check who is primary");
+    print();
+    print("\tsee also http://<mongod_host>:28017/_replSet for additional diagnostic info");
+}
 rs.slaveOk = function () { return db.getMongo().setSlaveOk(); }
 rs.status = function () { return db._adminCommand("replSetGetStatus"); }
 rs.isMaster = function () { return db.isMaster(); }