CVSROOT: /cvs/cluster
Module name: cluster
Changes by: [EMAIL PROTECTED] 2007-08-20 14:37:26
Modified files:
cman/cman_tool : main.c
cman/daemon : cnxman-private.h cnxman-socket.h commands.c
cman/lib : libcman.c libcman.h
Log message:
Add a "dirty" flag to cman to prevent active clusters merging with
one-another.
bz#251966
Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/cman_tool/main.c.diff?cvsroot=cluster&r1=1.53&r2=1.54
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/daemon/cnxman-private.h.diff?cvsroot=cluster&r1=1.26&r2=1.27
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/daemon/cnxman-socket.h.diff?cvsroot=cluster&r1=1.17&r2=1.18
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/daemon/commands.c.diff?cvsroot=cluster&r1=1.65&r2=1.66
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/lib/libcman.c.diff?cvsroot=cluster&r1=1.34&r2=1.35
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/cman/lib/libcman.h.diff?cvsroot=cluster&r1=1.30&r2=1.31
--- cluster/cman/cman_tool/main.c 2007/08/15 19:39:54 1.53
+++ cluster/cman/cman_tool/main.c 2007/08/20 14:37:26 1.54
@@ -235,6 +235,8 @@
printf(" Error");
if (einfo->ei_flags & CMAN_EXTRA_FLAG_DISALLOWED)
printf(" DisallowedNodes");
+ if (einfo->ei_flags & CMAN_EXTRA_FLAG_DIRTY)
+ printf(" Dirty");
printf(" \n");
printf("Ports Bound: ");
--- cluster/cman/daemon/cnxman-private.h 2006/11/03 15:07:52 1.26
+++ cluster/cman/daemon/cnxman-private.h 2007/08/20 14:37:26 1.27
@@ -18,8 +18,8 @@
/* Protocol Version triplet */
#define CNXMAN_MAJOR_VERSION 6
-#define CNXMAN_MINOR_VERSION 0
-#define CNXMAN_PATCH_VERSION 1
+#define CNXMAN_MINOR_VERSION 1
+#define CNXMAN_PATCH_VERSION 0
/* How we announce ourself in console events */
#define CMAN_NAME "CMAN"
@@ -147,11 +147,14 @@
NODE_FLAGS_FENCED - This node has been fenced since it last went
down.
NODE_FLAGS_FENCEDWHILEUP - This node was fenced manually (probably).
NODE_FLAGS_SEESDISALLOWED - Only set in a transition message
+ NODE_FLAGS_DIRTY - This node has internal state and must not join
+ a cluster that also has state.
*/
#define NODE_FLAGS_BEENDOWN 1
#define NODE_FLAGS_FENCED 2
#define NODE_FLAGS_FENCEDWHILEUP 4
#define NODE_FLAGS_SEESDISALLOWED 8
+#define NODE_FLAGS_DIRTY 16
/* There's one of these for each node in the cluster */
struct cluster_node {
--- cluster/cman/daemon/cnxman-socket.h 2006/10/05 07:48:33 1.17
+++ cluster/cman/daemon/cnxman-socket.h 2007/08/20 14:37:26 1.18
@@ -52,6 +52,7 @@
#define CMAN_CMD_GET_NODEADDRS 0x000000bf
#define CMAN_CMD_START_CONFCHG 0x000000c0
#define CMAN_CMD_STOP_CONFCHG 0x000000c1
+#define CMAN_CMD_SET_DIRTY 0x800000c2
#define CMAN_CMD_DATA 0x00000100
#define CMAN_CMD_BIND 0x00000101
@@ -167,6 +168,7 @@
#define CMAN_EXTRA_FLAG_ERROR 2
#define CMAN_EXTRA_FLAG_SHUTDOWN 4
#define CMAN_EXTRA_FLAG_UNCOUNTED 8
+#define CMAN_EXTRA_FLAG_DIRTY 16
struct cl_extra_info {
int node_state;
--- cluster/cman/daemon/commands.c 2007/06/26 09:09:13 1.65
+++ cluster/cman/daemon/commands.c 2007/08/20 14:37:26 1.66
@@ -502,6 +502,8 @@
einfo->flags |= CMAN_EXTRA_FLAG_SHUTDOWN;
if (uncounted)
einfo->flags |= CMAN_EXTRA_FLAG_UNCOUNTED;
+ if (us->flags & NODE_FLAGS_DIRTY)
+ einfo->flags |= CMAN_EXTRA_FLAG_DIRTY;
ptr = einfo->addresses;
for (i=0; i<num_interfaces; i++) {
@@ -1207,6 +1209,10 @@
err = 0;
break;
+ case CMAN_CMD_SET_DIRTY:
+ us->flags |= NODE_FLAGS_DIRTY;
+ break;
+
case CMAN_CMD_START_CONFCHG:
con->confchg = 1;
err = 0;
@@ -1706,7 +1712,27 @@
node = find_node_by_nodeid(nodeid);
assert(node);
- /* This is the killer. If the join_time of the node matches that
already stored AND
+ /* Newer nodes 6.1.0 onwards, set the DIRTY flag if they have state. If
the new node has been down
+ and has state then we mark it disallowed because we cannot merge
stateful nodes */
+ if (msg->flags & NODE_FLAGS_DIRTY && node->flags & NODE_FLAGS_BEENDOWN)
{
+ /* Don't duplicate messages */
+ if (node->state != NODESTATE_AISONLY) {
+ if (cluster_is_quorate) {
+ P_MEMB("Killing node %s because it has rejoined
the cluster with existing state", node->name);
+ log_printf(LOG_CRIT, "Killing node %s because
it has rejoined the cluster with existing state", node->name);
+ node->state = NODESTATE_AISONLY;
+ send_kill(nodeid, CLUSTER_KILL_REJOIN);
+ }
+ else {
+ P_MEMB("Node %s not joined to cman because it
has existing state", node->name);
+ log_printf(LOG_CRIT, "Node %s not joined to
cman because it has existing state", node->name);
+ node->state = NODESTATE_AISONLY;
+ }
+ }
+ return;
+ }
+
+ /* This is for older nodes. If the join_time of the node matches that
already stored AND
the node has been down, then we kill it as this must be a rejoin */
if (msg->join_time == node->cman_join_time && node->flags &
NODE_FLAGS_BEENDOWN) {
/* Don't duplicate messages */
--- cluster/cman/lib/libcman.c 2007/05/02 10:27:07 1.34
+++ cluster/cman/lib/libcman.c 2007/08/20 14:37:26 1.35
@@ -978,6 +978,14 @@
return info_call(h, CMAN_CMD_TRY_SHUTDOWN, &flags, sizeof(int), NULL,
0);
}
+int cman_set_dirty(cman_handle_t handle)
+{
+ struct cman_handle *h = (struct cman_handle *)handle;
+ VALIDATE_HANDLE(h);
+
+ return info_call(h, CMAN_CMD_SET_DIRTY, NULL, 0, NULL, 0);
+}
+
int cman_replyto_shutdown(cman_handle_t handle, int yesno)
{
struct cman_handle *h = (struct cman_handle *)handle;
--- cluster/cman/lib/libcman.h 2007/05/02 10:27:07 1.30
+++ cluster/cman/lib/libcman.h 2007/08/20 14:37:26 1.31
@@ -154,6 +154,7 @@
#define CMAN_EXTRA_FLAG_ERROR 2
#define CMAN_EXTRA_FLAG_SHUTDOWN 4
#define CMAN_EXTRA_FLAG_DISALLOWED 8
+#define CMAN_EXTRA_FLAG_DIRTY 16
typedef struct cman_extra_info {
int ei_node_state;
@@ -388,4 +389,12 @@
int cman_unregister_quorum_device(cman_handle_t handle);
int cman_poll_quorum_device(cman_handle_t handle, int isavailable);
+/*
+ * Sets the dirty bit inside cman. This indicates that the node has
+ * some internal 'state' (eg in a daemon, filesystem or lock manager)
+ * and cannot merge with another cluster that already has state.
+ * This cannot be reset.
+ */
+int cman_set_dirty(cman_handle_t handle);
+
#endif
