Rev 315: - don't try to send controls to dead nodes in http://samba.org/~tridge/ctdb

[tridge at samba.org]

------------------------------------------------------------
revno: 315
revision-id: tridge at samba.org-20070517132341-3syo9kztm9ckwpbs
parent: tridge at samba.org-20070517075402-8uppvsgamxi878c5
committer: Andrew Tridgell <tridge at samba.org>
branch nick: tridge
timestamp: Thu 2007-05-17 23:23:41 +1000
message:
  - don't try to send controls to dead nodes
  - use only connected nodes in a traverse
modified:
  common/ctdb.c                  ctdb.c-20061127094323-t50f58d65iaao5of-2
  common/ctdb_control.c          ctdb_control.c-20070426122724-j6gkpiofhbwdin63-1
  common/ctdb_recoverd.c         recoverd.c-20070503213540-bvxuyd9jm1f7ig90-1
  common/ctdb_traverse.c         ctdb_traverse.c-20070503021550-ztfs5rwx8jfm8qqx-1
  include/ctdb.h                 ctdb.h-20061117234101-o3qt14umlg9en8z0-11
  include/ctdb_private.h         ctdb_private.h-20061117234101-o3qt14umlg9en8z0-13
=== modified file 'common/ctdb.c'
--- a/common/ctdb.c	2007-05-12 09:54:40 +0000
+++ b/common/ctdb.c	2007-05-17 13:23:41 +0000
@@ -217,6 +217,21 @@
 	return ctdb->num_nodes;
 }
 
+/*
+  return the number of connected nodes
+*/
+uint32_t ctdb_get_num_connected_nodes(struct ctdb_context *ctdb)
+{
+	int i;
+	uint32_t count=0;
+	for (i=0;i<ctdb->vnn_map->size;i++) {
+		if (ctdb->nodes[i]->flags & NODE_FLAGS_CONNECTED) {
+			count++;
+		}
+	}
+	return count;
+}
+
 
 /*
   called by the transport layer when a packet comes in

=== modified file 'common/ctdb_control.c'
--- a/common/ctdb_control.c	2007-05-17 04:10:38 +0000
+++ b/common/ctdb_control.c	2007-05-17 13:23:41 +0000
@@ -395,6 +395,15 @@
 		return -1;
 	}
 
+	if (destnode != CTDB_BROADCAST_VNNMAP && destnode != CTDB_BROADCAST_VNNMAP && 
+	    (!ctdb_validate_vnn(ctdb, destnode) || 
+	     !(ctdb->nodes[destnode]->flags & NODE_FLAGS_CONNECTED))) {
+		if (!(flags & CTDB_CTRL_FLAG_NOREPLY)) {
+			callback(ctdb, -1, tdb_null, "ctdb_control to disconnected node", private_data);
+		}
+		return 0;
+	}
+
 	/* the state is made a child of private_data if possible. This means any reply
 	   will be discarded if the private_data goes away */
 	state = talloc(private_data?private_data:ctdb, struct ctdb_control_state);

=== modified file 'common/ctdb_recoverd.c'
--- a/common/ctdb_recoverd.c	2007-05-16 01:12:28 +0000
+++ b/common/ctdb_recoverd.c	2007-05-17 13:23:41 +0000
@@ -447,6 +447,8 @@
 		return -1;
 	}
 
+	/* send a message to all clients telling them that the cluster has been reconfigured */
+	ctdb_send_message(ctdb, CTDB_BROADCAST_ALL, CTDB_SRVID_RECONFIGURE, tdb_null);
 
 	DEBUG(0, (__location__ " Recovery complete\n"));
 	return 0;
@@ -465,7 +467,6 @@
 	uint64_t srvid;
 	
 	srvid = CTDB_SRVTYPE_RECOVERY;
-	srvid <<= 32;
 
 	emsg.vnn = vnn;
 
@@ -857,7 +858,6 @@
 
 	/* register a message port for recovery elections */
 	srvid = CTDB_SRVTYPE_RECOVERY;
-	srvid <<= 32;
 	ctdb_set_message_handler(ctdb, srvid, election_handler, NULL);
 
 	monitor_cluster(ctdb);

=== modified file 'common/ctdb_traverse.c'
--- a/common/ctdb_traverse.c	2007-05-10 07:43:45 +0000
+++ b/common/ctdb_traverse.c	2007-05-17 13:23:41 +0000
@@ -365,7 +365,7 @@
 
 	if (key.dsize == 0 && data.dsize == 0) {
 		state->null_count++;
-		if (state->null_count != ctdb_get_num_nodes(ctdb)) {
+		if (state->null_count != ctdb_get_num_connected_nodes(ctdb)) {
 			return 0;
 		}
 	}

=== modified file 'include/ctdb.h'
--- a/include/ctdb.h	2007-05-17 02:07:29 +0000
+++ b/include/ctdb.h	2007-05-17 13:23:41 +0000
@@ -56,10 +56,16 @@
    a message handler ID meaning "give me all messages"
  */
 #define CTDB_SRVID_ALL (~(uint64_t)0)
+
 /*
   srvid type : RECOVERY
 */
-#define CTDB_SRVTYPE_RECOVERY	0x64766372
+#define CTDB_SRVTYPE_RECOVERY	0xF100000000000000LL
+
+/* 
+   a message handler ID meaning that the cluster has been reconfigured
+ */
+#define CTDB_SRVID_RECONFIGURE 0xF200000000000000LL
 
 struct event_context;
 

=== modified file 'include/ctdb_private.h'
--- a/include/ctdb_private.h	2007-05-17 04:10:38 +0000
+++ b/include/ctdb_private.h	2007-05-17 13:23:41 +0000
@@ -802,4 +802,6 @@
 
 int ctdb_start_recoverd(struct ctdb_context *ctdb);
 
+uint32_t ctdb_get_num_connected_nodes(struct ctdb_context *ctdb);
+
 #endif