we only clone one database per pass, even if a lot need done. This helps us

avoid overflowing the master's transaction log by doing too much work before going
back to read more transactions. (Imagine a scenario of slave startup where we try to
clone 100 databases in one pass.)

4 files changed, 19 insertions, 6 deletions

diff --git a/db/pdfile.cpp b/db/pdfile.cpp
index d2fb10b4140..122f54e2b17 100644
--- a/db/pdfile.cpp
+++ b/db/pdfile.cpp
@@ -402,7 +402,7 @@ int initialExtentSize(int len) {
 		sz = 1000000000;
 	int z = ((int)sz) & 0xffffff00;
 	assert( z > len );
-	log() << "initialExtentSize(" << len << ") returns " << z << endl;
+	DEV log() << "initialExtentSize(" << len << ") returns " << z << endl;
 	return z;
 }
 
@@ -478,7 +478,7 @@ void PhysicalDataFile::open(int fn, const char *filename) {
 
 	assert(fn == fileNo);
 	header = (PDFHeader *) mmf.map(filename, length);
-	assert(header);
+	uassert("can't map file memory", header);
 	header->init(fileNo, length);
 }
 
diff --git a/db/query.cpp b/db/query.cpp
index 4f99cfa0a85..e4df9643ab5 100644
--- a/db/query.cpp
+++ b/db/query.cpp
@@ -641,7 +641,7 @@ bool _runCommands(const char *ns, JSObj& jsobj, stringstream& ss, BufBuilder &b,
 					ok = false;
 				} else { 
 					dropDatabase(ns);
-//finish                    logOp("c", ns, jsobj);
+					logOp("c", ns, jsobj);
 					ok = true;
 				}
 			}
diff --git a/db/repl.cpp b/db/repl.cpp
index 13789376b1f..12722b37377 100644
--- a/db/repl.cpp
+++ b/db/repl.cpp
@@ -68,7 +68,7 @@ int test2() {
 
 /* --------------------------------------------------------------*/
 
-Source::Source(JSObj o) {
+Source::Source(JSObj o) : nClonedThisPass(0) {
 	only = o.getStringField("only");
 	hostName = o.getStringField("host");
 	sourceName = o.getStringField("source");
@@ -210,13 +210,23 @@ void Source::applyOperation(JSObj& op) {
 	if( !only.empty() && only != clientName )
 		return;
 
-	dblock lk;
+	bool newDb = dbs.count(clientName) == 0;
+	if( newDb && nClonedThisPass ) { 
+		/* we only clone one database per pass, even if a lot need done.  This helps us 
+		   avoid overflowing the master's transaction log by doing too much work before going 
+		   back to read more transactions. (Imagine a scenario of slave startup where we try to 
+		   clone 100 databases in one pass.)
+		*/
+		return;
+	}
 
+	dblock lk;
 	setClientTempNs(ns);
 
 	if( client->justCreated || /* datafiles were missing.  so we need everything, no matter what sources object says */
-	    !dbs.count(client->name) ) /* if not in dbs, we've never synced this database before, so we need everything */
+	    newDb ) /* if not in dbs, we've never synced this database before, so we need everything */
 	{
+		nClonedThisPass++;
 		resync(client->name);
 		client->justCreated = false;
 	}
@@ -360,6 +370,7 @@ void Source::pullOpLog() {
 */
 bool Source::sync() { 
 	log() << "pull: " << sourceName << '@' << hostName << endl;
+	nClonedThisPass = 0;
 
 	if( (string("localhost") == hostName || string("127.0.0.1") == hostName) && port == DBPort ) { 
         log() << "pull:   can't sync from self (localhost). sources configuration may be wrong." << endl;
diff --git a/db/repl.h b/db/repl.h
index ee343f547d2..d3708f56860 100644
--- a/db/repl.h
+++ b/db/repl.h
@@ -106,6 +106,8 @@ public:
 	*/
 	set<string> dbs;
 
+	int nClonedThisPass;
+
 	static void loadAll(vector<Source*>&);
 	static void cleanup(vector<Source*>&);
 	Source(JSObj);