Rev 346: make sure we find out about new nodes as fast as possible in http://samba.org/~tridge/ctdb

[tridge at samba.org]

------------------------------------------------------------
revno: 346
revision-id: tridge at samba.org-20070525120745-40ik47s270sgbx07
parent: tridge at samba.org-20070525112726-m22jkf53zf4xkfpc
committer: Andrew Tridgell <tridge at samba.org>
branch nick: tridge
timestamp: Fri 2007-05-25 22:07:45 +1000
message:
  make sure we find out about new nodes as fast as possible
modified:
  tcp/ctdb_tcp.h                 ctdb_tcp.h-20061127103747-l8xeniwiapbydehq-3
  tcp/tcp_connect.c              tcp_connect.c-20061128004937-x70q1cu5xzg5g2tm-1
=== modified file 'tcp/ctdb_tcp.h'
--- a/tcp/ctdb_tcp.h	2007-04-13 10:38:24 +0000
+++ b/tcp/ctdb_tcp.h	2007-05-25 12:07:45 +0000
@@ -39,6 +39,8 @@
 struct ctdb_tcp_node {
 	int fd;
 	struct ctdb_queue *queue;
+	struct fd_event *connect_fde;
+	struct timed_event *connect_te;
 };
 
 

=== modified file 'tcp/tcp_connect.c'
--- a/tcp/tcp_connect.c	2007-05-15 08:40:56 +0000
+++ b/tcp/tcp_connect.c	2007-05-25 12:07:45 +0000
@@ -50,6 +50,7 @@
 	/* start a new connect cycle to try to re-establish the
 	   link */
 	ctdb_queue_set_fd(tnode->queue, -1);
+	tnode->fd = -1;
 	event_add_timed(node->ctdb->ev, node, timeval_zero(), 
 			ctdb_tcp_node_connect, node);
 }
@@ -69,6 +70,9 @@
 	socklen_t len = sizeof(error);
 	int one = 1;
 
+	talloc_free(tnode->connect_te);
+	tnode->connect_te = NULL;
+
 	if (getsockopt(tnode->fd, SOL_SOCKET, SO_ERROR, &error, &len) != 0 ||
 	    error != 0) {
 		talloc_free(fde);
@@ -120,6 +124,13 @@
         struct sockaddr_in sock_in;
         struct sockaddr_in sock_out;
 
+	if (tnode->fd != -1) {
+		talloc_free(tnode->connect_fde);
+		tnode->connect_fde = NULL;
+		close(tnode->fd);
+		tnode->fd = -1;
+	}
+
 	tnode->fd = socket(PF_INET, SOCK_STREAM, IPPROTO_TCP);
 
 	set_nonblocking(tnode->fd);
@@ -163,8 +174,15 @@
 	}
 
 	/* non-blocking connect - wait for write event */
-	event_add_fd(node->ctdb->ev, node, tnode->fd, EVENT_FD_WRITE|EVENT_FD_READ, 
-		     ctdb_node_connect_write, node);
+	tnode->connect_fde = event_add_fd(node->ctdb->ev, node, tnode->fd, 
+					  EVENT_FD_WRITE|EVENT_FD_READ, 
+					  ctdb_node_connect_write, node);
+
+	/* don't give it long to connect - retry in one second. This ensures
+	   that we find a node is up quickly (tcp normally backs off a syn reply
+	   delay by quite a lot) */
+	tnode->connect_te = event_add_timed(ctdb->ev, node, timeval_current_ofs(1, 0), 
+					    ctdb_tcp_node_connect, node);
 }
 
 /*