SERVER-9481 Fixes fluctuating between ROLLBACK and SECONDARY when ROLLBACK is impossible

Only leave ROLLBACK when successful and go to FATAL when ROLLBACK is impossible
because we are not willing to rollback so much data

Signed-off-by: Matt Dannenberg <matt.dannenberg@10gen.com>

2 files changed, 12 insertions, 5 deletions

diff --git a/src/mongo/db/repl/rs_rollback.cpp b/src/mongo/db/repl/rs_rollback.cpp
index 8c8e89c20c6..a94e4603f60 100644
--- a/src/mongo/db/repl/rs_rollback.cpp
+++ b/src/mongo/db/repl/rs_rollback.cpp
@@ -68,7 +68,12 @@ namespace mongo {
 
     class rsfatal : public std::exception {
     public:
-        virtual const char* what() const throw() { return "replica set fatal exception"; }
+        rsfatal(const char* m = "replica set fatal exception") {
+            msg = m;
+        }
+        virtual const char* what() const throw() { return msg; }
+    private:
+        const char *msg;
     };
 
     struct DocID {
@@ -230,7 +235,7 @@ namespace mongo {
             log() << "replSet info rollback diff in end of log times: " << diff << " seconds" << rsLog;
             if( diff > 1800 ) {
                 log() << "replSet rollback too long a time period for a rollback." << rsLog;
-                throw "error not willing to roll back more than 30 minutes of data";
+                throw rsfatal("rollback error: not willing to roll back more than 30 minutes of data");
             }
         }
 
@@ -641,7 +646,8 @@ namespace mongo {
                 sethbmsg(string("rollback 2 error ") + p);
                 return 10;
             }
-            catch( rsfatal& ) {
+            catch( rsfatal& e ) {
+                sethbmsg(string(e.what()));
                 _fatal();
                 return 2;
             }
diff --git a/src/mongo/db/repl/rs_sync.cpp b/src/mongo/db/repl/rs_sync.cpp
index f1c531b1c7d..5ea66deef5b 100644
--- a/src/mongo/db/repl/rs_sync.cpp
+++ b/src/mongo/db/repl/rs_sync.cpp
@@ -587,8 +587,9 @@ namespace replset {
         lock rsLock( this );
         Lock::GlobalWrite writeLock;
 
-        // make sure we're not primary, secondary, or fatal already
-        if (box.getState().primary() || box.getState().secondary() || box.getState().fatal()) {
+        // make sure we're not primary, secondary, rollback, or fatal already
+        if (box.getState().primary() || box.getState().secondary() ||
+            box.getState().rollback() || box.getState().fatal()) {
             return false;
         }