started adding a cleaner daemon finish method

(This used to be ctdb commit 5ef0cd83d7f24616dad85cece485b770376ecd45)

6 files changed, 35 insertions, 11 deletions

diff --git a/ctdb/common/ctdb.c b/ctdb/common/ctdb.c
index 60e1e6b90b..a086fc4cfd 100644
--- a/ctdb/common/ctdb.c
+++ b/ctdb/common/ctdb.c
@@ -255,6 +255,10 @@ void ctdb_recv_pkt(struct ctdb_context *ctdb, uint8_t *data, uint32_t length)
 		ctdb_request_message(ctdb, hdr);
 		break;
 
+	case CTDB_REQ_FINISHED:
+		ctdb_request_finished(ctdb, hdr);
+		break;
+
 	default:
 		DEBUG(0,("%s: Packet with unknown operation %d\n", 
 			 __location__, hdr->operation));
diff --git a/ctdb/common/ctdb_client.c b/ctdb/common/ctdb_client.c
index aca243eda1..4e5e7701cc 100644
--- a/ctdb/common/ctdb_client.c
+++ b/ctdb/common/ctdb_client.c
@@ -434,6 +434,11 @@ void ctdb_connect_wait(struct ctdb_context *ctdb)
 	r.hdr.operation = CTDB_REQ_CONNECT_WAIT;
 
 	DEBUG(3,("ctdb_connect_wait: sending to ctdbd\n"));
+
+	/* if the domain socket is not yet open, open it */
+	if (ctdb->daemon.sd==-1) {
+		ux_socket_connect(ctdb);
+	}
 	
 	res = ctdb_queue_send(ctdb->daemon.queue, (uint8_t *)&r.hdr, r.hdr.length);
 	if (res != 0) {
diff --git a/ctdb/common/ctdb_daemon.c b/ctdb/common/ctdb_daemon.c
index a4bb36d482..1e85440efd 100644
--- a/ctdb/common/ctdb_daemon.c
+++ b/ctdb/common/ctdb_daemon.c
@@ -611,3 +611,15 @@ void *ctdbd_allocate_pkt(struct ctdb_context *ctdb, size_t len)
 	return talloc_size(ctdb, size);
 }
 
+/*
+  called when a CTDB_REQ_FINISHED packet comes in
+*/
+void ctdb_request_finished(struct ctdb_context *ctdb, struct ctdb_req_header *hdr)
+{
+	ctdb->num_finished++;
+	if (ctdb->num_finished == ctdb->num_nodes) {
+		/* all daemons have requested to finish - we now exit */
+		DEBUG(1,("All daemons finished - exiting\n"));
+		_exit(0);
+	}
+}
diff --git a/ctdb/include/ctdb_private.h b/ctdb/include/ctdb_private.h
index 6736ee4ab7..bf6b96e635 100644
--- a/ctdb/include/ctdb_private.h
+++ b/ctdb/include/ctdb_private.h
@@ -117,6 +117,7 @@ struct ctdb_context {
 	uint32_t vnn; /* our own vnn */
 	uint32_t num_nodes;
 	uint32_t num_connected;
+	uint32_t num_finished;
 	unsigned flags;
 	struct idr_context *idr;
 	struct ctdb_node **nodes; /* array of nodes in the cluster - indexed by vnn */
@@ -213,6 +214,7 @@ enum ctdb_operation {
 	CTDB_REPLY_DMASTER      = 4,
 	CTDB_REPLY_ERROR        = 5,
 	CTDB_REQ_MESSAGE        = 6,
+	CTDB_REQ_FINISHED       = 7,
 	
 	/* only used on the domain socket */
 	CTDB_REQ_REGISTER       = 1000,     
@@ -294,6 +296,10 @@ struct ctdb_req_message {
 	uint8_t data[1];
 };
 
+struct ctdb_req_finished {
+	struct ctdb_req_header hdr;
+};
+
 struct ctdb_req_connect_wait {
 	struct ctdb_req_header hdr;
 };
diff --git a/ctdb/tests/ctdb_test.c b/ctdb/tests/ctdb_test.c
index dea9066eb2..e051305725 100644
--- a/ctdb/tests/ctdb_test.c
+++ b/ctdb/tests/ctdb_test.c
@@ -133,6 +133,8 @@ int main(int argc, const char *argv[])
 	/* start the protocol running */
 	ret = ctdb_start(ctdb);
 
+	ctdb_connect_wait(ctdb);
+
 	ZERO_STRUCT(call);
 	call.key.dptr = discard_const("test");
 	call.key.dsize = strlen("test")+1;
@@ -171,11 +173,6 @@ int main(int argc, const char *argv[])
 	/* go into a wait loop to allow other nodes to complete */
 	ctdb_wait_loop(ctdb);
 
-	/*talloc_report_full(ctdb, stdout);*/
-
-/* sleep for a while so that our daemon will remaining alive for the other nodes in the cluster */
-sleep(10);
-
 	/* shut it down */
 	talloc_free(ctdb);
 	return 0;
diff --git a/ctdb/tests/test.sh b/ctdb/tests/test.sh
index f119ad5b1b..b1ed1d20d7 100755
--- a/ctdb/tests/test.sh
+++ b/ctdb/tests/test.sh
@@ -4,17 +4,17 @@ killall -q ctdb_test
 
 
 echo "Trying 2 nodes"
-bin/ctdb_test --nlist tests/nodes.txt --listen 127.0.0.1:9001 &
-bin/ctdb_test --nlist tests/nodes.txt --listen 127.0.0.2:9001
+$VALGRIND bin/ctdb_test --nlist tests/nodes.txt --listen 127.0.0.1:9001 &
+$VALGRIND bin/ctdb_test --nlist tests/nodes.txt --listen 127.0.0.2:9001
 
 sleep 3
 killall ctdb_test
 
 echo "Trying 4 nodes"
-bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.1:9001 &
-bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.2:9001 &
-bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.3:9001 &
-bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.4:9001
+$VALGRIND bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.1:9001 &
+$VALGRIND bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.2:9001 &
+$VALGRIND bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.3:9001 &
+$VALGRIND bin/ctdb_test --nlist tests/4nodes.txt --listen 127.0.0.4:9001
 sleep 3
 
 killall ctdb_test