Pre-approval for heartbeat testing-proposed-updates upload (2.1.3-6lenny1)
Hi,
I would like to upload a(nother) fresh version of heartbeat to fix
a fairly severe bug. The fix has been in unstable since 2.1.3-7 and
was included upstream at around the same time - it seems to be well tested.
heartbeat (2.1.3-6lenny1) testing-proposed-updates; urgency=low
* dopd: fix basic failover; fix hb message corruption by
* fprintf(stderr)
Patch: fix-basic-failover-fix-hb-message-corruption-by-fprintf.patch
Upstream-Status: commit 47f60bebe7b25abd88ea7b5488e66dfe187416ae
"dopd: fix basic failover; fix hb message corruption by
fprintf(stderr)"
(closes: #486071)
-- Simon Horman <horms@debian.org> Fri, 13 Feb 2009 10:30:31 +1100
The patch is below
--
Simon Horman
VA Linux Systems Japan K.K., Sydney, Australia Satellite Office
H: www.vergenet.net/~horms/ W: www.valinux.co.jp/en
# HG changeset patch
# User Rasto Levrinc <rasto@linbit.com>
# Date 1206539836 -3600
# Node ID 47f60bebe7b25abd88ea7b5488e66dfe187416ae
# Parent 17c0cf487322287d0689a036c32f21b900ce5a80
dopd: fix basic failover; fix hb message corruption by fprintf(stderr)
check_drbd_peer() used to return FALSE for "node name not in node list",
so drbd-peer-outdater returned "invalid nodename".
Then the semantic changed, and check_drbd_peer learned about "dead" peers
and returned FALSE for them as well. Which made basic failover impossible :(
The return code was now changed to "peer unreachable" for a dead peer.
And even for nodes which really are not in the host list (and thus could be
classified as invalide), because, after all, thats what they are.
unreachable.
Node name comparison needs to be case insensitive; fixed.
During testing with 15 concurrent drbd resources several dopd crashes have been
observed, which after some debugging turned out to be simply a wrong assumption
about the global availability of stderr: some fprintf(stderr, "debug message")
had accidentally used the heartbeat communication channel file descriptor,
which seriously confused the comm layer.
All those fprintfs have now been changed to use cl_log.
diff -r 17c0cf487322 -r 47f60bebe7b2 contrib/drbd-outdate-peer/dopd.c
--- a/contrib/drbd-outdate-peer/dopd.c Mon Mar 24 16:14:12 2008 +0100
+++ b/contrib/drbd-outdate-peer/dopd.c Wed Mar 26 14:57:16 2008 +0100
@@ -202,14 +202,17 @@
}
/* check_drbd_peer()
- * walk the nodes and return TRUE if peer is not this node and it exists.
+ * walk the nodes and return
+ * FALSE if peer is not found, not a "normal" node, or "dead"
+ * (no point in trying to reach those nodes).
+ * TRUE if peer is found to be alive and "normal".
*/
gboolean
check_drbd_peer(const char *drbd_peer)
{
const char *node;
gboolean found = FALSE;
- if (!strcmp(drbd_peer, node_name)) {
+ if (!strcasecmp(drbd_peer, node_name)) {
cl_log(LOG_WARNING, "drbd peer node %s is me!\n", drbd_peer);
return FALSE;
}
@@ -306,9 +309,9 @@
} else
pthread_mutex_unlock(&conn_mutex);
} else {
- /* wrong peer was specified,
- send return code 20 to the client */
- send_to_client(curr_client, "20");
+ /* peer "dead" or not in node list.
+ * return "peer could not be reached" */
+ send_to_client(curr_client, "5");
}
ha_msg_del(msg);
diff -r 17c0cf487322 -r 47f60bebe7b2 contrib/drbd-outdate-peer/drbd-peer-outdater.c
--- a/contrib/drbd-outdate-peer/drbd-peer-outdater.c Mon Mar 24 16:14:12 2008 +0100
+++ b/contrib/drbd-outdate-peer/drbd-peer-outdater.c Wed Mar 26 14:57:16 2008 +0100
@@ -76,7 +76,7 @@
msg = msgfromIPC_noauth(server);
if (!msg) {
- fprintf(stderr, "no message from server or other "
+ cl_log(LOG_WARNING, "no message from server or other "
"instance is running\n");
if (client->mainloop != NULL &&
g_main_is_running(client->mainloop))
@@ -92,7 +92,7 @@
errno = 0;
rc = strtol(rc_string, &ep, 10);
if (errno != 0 || *ep != EOS) {
- fprintf(stderr, "unknown message: %s from server", rc_string);
+ cl_log(LOG_WARNING, "unknown message: %s from server", rc_string);
client->rc = 20; /* "officially undefined", unspecified error */
ha_msg_del(msg);
if (client->mainloop != NULL &&
@@ -124,7 +124,7 @@
outdater_timeout_dispatch(gpointer user_data)
{
dop_client_t *client = (dop_client_t *)user_data;
- fprintf(stderr, "error: could not connect to dopd after %i seconds"
+ cl_log(LOG_WARNING, "error: could not connect to dopd after %i seconds"
": timeout reached\n", client->timeout);
if (client->mainloop != NULL && g_main_is_running(client->mainloop))
g_main_quit(client->mainloop);
@@ -255,7 +255,7 @@
(gpointer)new_client, &ipc_server);
if (ipc_server == NULL) {
- fprintf(stderr, "Could not connect to "T_OUTDATER" channel\n");
+ cl_log(LOG_WARNING, "Could not connect to "T_OUTDATER" channel\n");
dop_exit(new_client); /* unreachable */
}
@@ -267,7 +267,7 @@
ha_msg_add(update, F_OUTDATER_RES, drbd_resource);
if (msg2ipcchan(update, ipc_server) != HA_OK) {
- fprintf(stderr, "Could not send message\n");
+ cl_log(LOG_WARNING, "Could not send message\n");
dop_exit(new_client);
}
Reply to: