Rev 146: started adding a cleaner daemon finish method in http://samba.org/~tridge/ctdb

[tridge at samba.org]

------------------------------------------------------------
revno: 146
revision-id: tridge at samba.org-20070418015554-f1b0faceed62a64b
parent: tridge at samba.org-20070418013316-60979fe73163efe3
committer: Andrew Tridgell <tridge at samba.org>
branch nick: tridge
timestamp: Wed 2007-04-18 11:55:54 +1000
message:
  started adding a cleaner daemon finish method
modified:
  common/ctdb.c                  ctdb.c-20061127094323-t50f58d65iaao5of-2
  common/ctdb_client.c           ctdb_client.c-20070411010216-3kd8v37k61steeya-1
  common/ctdb_daemon.c           ctdb_daemon.c-20070409200331-3el1kqgdb9m4ib0g-1
  include/ctdb_private.h         ctdb_private.h-20061117234101-o3qt14umlg9en8z0-13
  tests/ctdb_test.c              ctdb_test.c-20061117234101-o3qt14umlg9en8z0-16
  tests/test.sh                  test.sh-20061128065333-cla80zaxd9eb4o4a-2
=== modified file 'common/ctdb.c'
--- a/common/ctdb.c	2007-04-18 01:20:24 +0000
+++ b/common/ctdb.c	2007-04-18 01:55:54 +0000
@@ -255,6 +255,10 @@
 		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));

=== modified file 'common/ctdb_client.c'
--- a/common/ctdb_client.c	2007-04-18 01:20:24 +0000
+++ b/common/ctdb_client.c	2007-04-18 01:55:54 +0000
@@ -434,6 +434,11 @@
 	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) {

=== modified file 'common/ctdb_daemon.c'
--- a/common/ctdb_daemon.c	2007-04-18 01:33:16 +0000
+++ b/common/ctdb_daemon.c	2007-04-18 01:55:54 +0000
@@ -611,3 +611,15 @@
 	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);
+	}
+}

=== modified file 'include/ctdb_private.h'
--- a/include/ctdb_private.h	2007-04-18 01:20:24 +0000
+++ b/include/ctdb_private.h	2007-04-18 01:55:54 +0000
@@ -117,6 +117,7 @@
 	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 @@
 	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 @@
 	uint8_t data[1];
 };
 
+struct ctdb_req_finished {
+	struct ctdb_req_header hdr;
+};
+
 struct ctdb_req_connect_wait {
 	struct ctdb_req_header hdr;
 };

=== modified file 'tests/ctdb_test.c'
--- a/tests/ctdb_test.c	2007-04-17 12:16:50 +0000
+++ b/tests/ctdb_test.c	2007-04-18 01:55:54 +0000
@@ -133,6 +133,8 @@
 	/* 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 @@
 	/* 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;

=== modified file 'tests/test.sh'
--- a/tests/test.sh	2007-04-17 05:01:42 +0000
+++ b/tests/test.sh	2007-04-18 01:55:54 +0000
@@ -4,17 +4,17 @@
 
 
 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