move 2nd occurence of same collnum_t collection id

on the same shard to the trash/ subdir.
put call to syncParmsWithHost0 in a sleep loop in case
host #0 has error, although the timeout is really high.

Collectiondb.cpp

@@ -252,6 +252,26 @@ bool Collectiondb::addExistingColl ( char *coll, collnum_t collnum ) {
 		char *xx=NULL;*xx=0;
 	}
 
+	// also try by #, i've seen this happen too
+	CollectionRec *ocr = getRec ( i );
+	if ( ocr ) {
+		g_errno = EEXIST;
+		log("admin: Collection id %i is in use already by "
+		    "%s, so we can not add %s. moving %s to trash."
+		    ,(int)i,ocr->m_coll,coll,coll);
+		SafeBuf cmd;
+		int64_t now = gettimeofdayInMilliseconds();
+		cmd.safePrintf ( "mv coll.%s.%i trash/coll.%s.%i.%"UINT64
+				 , coll
+				 ,(int)i
+				 , coll
+				 ,(int)i
+				 , now );
+		//log("admin: %s",cmd.getBufStart());
+		gbsystem ( cmd.getBufStart() );
+		return true;
+	}
+
 	// create the record in memory
 	CollectionRec *cr = new (CollectionRec);
 	if ( ! cr ) 

Parms.cpp

@@ -973,6 +973,7 @@ class Checksum {
 Parms::Parms ( ) {
 	m_isDefaultLoaded = false;
 	m_inSyncWithHost0 = false;
+	m_triedToSync     = false;
 }
 
 void Parms::detachSafeBufs ( CollectionRec *cr ) {
@@ -21736,20 +21737,37 @@ void handleRequest3f ( UdpSlot *slot , int32_t niceness ) {
 //    have with ETRYAGAIN in Msg4.cpp
 
 
+void tryToSyncWrapper ( int fd , void *state ) {
+	g_parms.syncParmsWithHost0();
+}
+
 // host #0 just sends back an empty reply, but it will hit us with
 // 0x3f parmlist requests. that way it uses the same mechanism and can
 // guarantee ordering of the parm update requests
 void gotReplyFromHost0Wrapper ( void *state , UdpSlot *slot ) {
 	// ignore his reply unless error?
-	if ( g_errno )
-		log("parms: got error syncing with host 0: %s",
+	if ( g_errno ) {
+		log("parms: got error syncing with host 0: %s. Retrying.",
 		    mstrerror(g_errno));
+		// re-try it!
+		g_parms.m_triedToSync = false;
+	}
+	else {
+		log("parms: synced with host #0");
+		// do not re-call
+		g_loop.unregisterSleepCallback(NULL,tryToSyncWrapper);
+	}
+
 	g_errno = 0;
 }
-	
+
 // returns false and sets g_errno on error, true otherwise
 bool Parms::syncParmsWithHost0 ( ) {
 
+	if ( m_triedToSync ) return true;
+
+	m_triedToSync = true;
+
 	m_inSyncWithHost0 = false;
 
 	// dont sync with ourselves
@@ -21782,6 +21800,8 @@ bool Parms::syncParmsWithHost0 ( ) {
 
 	Host *h = g_hostdb.getHost(0);
 
+	log("parms: trying to sync with host #0");
+
 	// . send it off. use 3e i guess
 	// . host #0 will reply using msg4 really
 	// . msg4 guarantees ordering of requests

Parms.h

@@ -516,6 +516,7 @@ class Parms {
 	//
 
 	bool m_inSyncWithHost0;
+	bool m_triedToSync;
 
 	bool m_isDefaultLoaded;
 

main.cpp

@@ -373,6 +373,7 @@ extern void resetQuery         ( );
 extern void resetStopWords     ( );
 extern void resetUnicode       ( );
 
+extern void tryToSyncWrapper ( int fd , void *state ) ;
 
 #if 0
 void stack_test();
@@ -3855,7 +3856,9 @@ int main2 ( int argc , char *argv[] ) {
 	     ! g_loop.registerSleepCallback(2000,(void *)1,runSEOQueryLoop))
 		log("db: Failed to register seo query loop");
 
-
+	// try to sync parms (and collection recs) with host 0
+	if ( ! g_loop.registerSleepCallback(1000,NULL,tryToSyncWrapper,0))
+		return false;
 
 	//if( !g_loop.registerSleepCallback(2000,(void *)1,controlDumpTopDocs) )
 	//	log("db: Failed to init dump TopDocs sleep callback.");
@@ -3873,11 +3876,11 @@ int main2 ( int argc , char *argv[] ) {
 	//msg3e.checkForNewParms();
 
 	// this stuff is similar to alden's msg3e but will sync collections
-	// that were added/deleted
-	if ( ! g_parms.syncParmsWithHost0() ) {
-		log("parms: error syncing parms: %s",mstrerror(g_errno));
-		return 0;
-	}
+	// that were added/deletede
+	//if ( ! g_parms.syncParmsWithHost0() ) {
+	//	log("parms: error syncing parms: %s",mstrerror(g_errno));
+	//	return 0;
+	//}
 
 
 	if(g_recoveryMode) {