[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Cluster-devel] cluster/rgmanager src/daemons/rg_forward.c src ...



CVSROOT:	/cvs/cluster
Module name:	cluster
Branch: 	RHEL4
Changes by:	lhh sourceware org	2006-12-13 18:19:57

Modified files:
	rgmanager/src/daemons: rg_forward.c resrules.c rg_thread.c 
	                       reslist.c rg_state.c main.c 
	rgmanager/init.d: rgmanager 
	rgmanager/src/clulib: vft.c 
	rgmanager/src/utils: clusvcadm.c clustat.c 
	rgmanager/src/resources: ip.sh 
	rgmanager/include: resgroup.h 

Log message:
	Fix #193603 (part 2; return error code if rgmanager start fails), #210455, #212110, #216774, #212634/#218112.
	Fix bug in ip.sh allowing start of the IP if the link was down, preventing failover (linux-cluster reported).

Patches:
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/rg_forward.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.2.2.1&r2=1.2.2.2
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/resrules.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.9.2.4&r2=1.9.2.5
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/rg_thread.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.7.2.7&r2=1.7.2.8
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/reslist.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.6.2.6&r2=1.6.2.7
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/rg_state.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.4.2.16&r2=1.4.2.17
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/daemons/main.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.9.2.20&r2=1.9.2.21
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/init.d/rgmanager.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.3.2.3&r2=1.3.2.4
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/clulib/vft.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.7.2.6&r2=1.7.2.7
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/utils/clusvcadm.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.2.2.6&r2=1.2.2.7
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/utils/clustat.c.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.5.2.14&r2=1.5.2.15
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/src/resources/ip.sh.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.5.2.15&r2=1.5.2.16
http://sourceware.org/cgi-bin/cvsweb.cgi/cluster/rgmanager/include/resgroup.h.diff?cvsroot=cluster&only_with_tag=RHEL4&r1=1.3.2.7&r2=1.3.2.8

--- cluster/rgmanager/src/daemons/rg_forward.c	2006/01/20 16:25:25	1.2.2.1
+++ cluster/rgmanager/src/daemons/rg_forward.c	2006/12/13 18:19:56	1.2.2.2
@@ -97,7 +97,7 @@
 	msg_close(fd);
 
 	swab_SmMessageSt(&msg);
-	send_response(msg.sm_data.d_ret, req);
+	send_response(msg.sm_data.d_ret, req->rr_target, req);
 
 	rq_free(req);
 
--- cluster/rgmanager/src/daemons/resrules.c	2006/10/20 20:57:19	1.9.2.4
+++ cluster/rgmanager/src/daemons/resrules.c	2006/12/13 18:19:56	1.9.2.5
@@ -248,17 +248,36 @@
 }
 
 
+/**
+ * Store a resource action
+ * @param actsp		Action array; may be modified and returned!
+ * @param name		Name of the action
+ * @param depth		Resource depth (status/monitor; -1 means *ALL LEVELS*
+ * 			... this means that only the highest-level check depth
+ * 			will ever be performed!)
+ * @param timeout	Timeout (not used)
+ * @param interval	Time interval for status/monitor
+ * @return		0 on success, -1 on failure
+ * 
+ */
 int
 store_action(resource_act_t **actsp, char *name, int depth,
 	     int timeout, int interval)
 {
-	int x = 0;
+	int x = 0, replace = 0;
 	resource_act_t *acts = *actsp;
 
 	if (!name)
 		return -1;
+	
+	if (depth < 0 && timeout < 0 && interval < 0)
+		return -1;
 
 	if (!acts) {
+		/* Can't create with anything < 0 */
+		if (depth < 0 || timeout < 0 || interval < 0)
+			return -1;
+		
 		acts = malloc(sizeof(resource_act_t) * 2);
 		if (!acts)
 			return -1;
@@ -274,17 +293,38 @@
 
 	for (x = 0; acts[x].ra_name; x++) {
 		if (!strcmp(acts[x].ra_name, name) &&
-		    depth == acts[x].ra_depth) {
-			printf("Skipping duplicate action/depth %s/%d\n",
-			       name, depth);
-			return -1;
+		    (depth == acts[x].ra_depth || depth == -1)) {
+			printf("Replacing action '%s' depth %d: ",
+			       name, acts[x].ra_depth);
+			if (timeout >= 0) {
+				printf("timeout: %d->%d ",
+				       (int)acts[x].ra_timeout,
+				       (int)timeout);
+				acts[x].ra_timeout = timeout;
+			}
+			if (interval >= 0) {
+				printf("interval: %d->%d",
+				       (int)acts[x].ra_interval,
+				       (int)interval);
+				acts[x].ra_interval = interval;
+			}
+			printf("\n");
+			replace = 1;
 		}
 	}
+	
+	if (replace)
+		/* If we replaced something, we're done */
+		return 1;
+	
+	/* Can't create with anything < 0 */
+	if (depth < 0 || timeout < 0 || interval < 0)
+		return -1;
 
 	acts = realloc(acts, sizeof(resource_act_t) * (x+2));
 	if (!acts)
 		return -1;
-
+	
 	acts[x].ra_name = name;
 	acts[x].ra_depth = depth;
 	acts[x].ra_timeout = timeout;
@@ -297,6 +337,7 @@
 }
 
 
+
 void
 _get_actions(xmlDocPtr doc, xmlXPathContextPtr ctx, char *base,
 		 resource_rule_t *rr)
@@ -324,8 +365,8 @@
 		ret = xpath_get_one(doc, ctx, xpath);
 		if (ret) {
 			timeout = expand_time(ret);
-			if (interval < 0)
-				interval = 0;
+			if (timeout < 0)
+				timeout = 0;
 			free(ret);
 		}
 
@@ -352,9 +393,8 @@
 		}
 
 		if (store_action(&rr->rr_actions, act, depth, timeout,
-				 interval) < 0)
+				 interval) != 0)
 			free(act);
-		
 	} while (1);
 
 
@@ -946,6 +986,14 @@
 		type = xpath_get_one(doc, ctx, base);
 		if (!type)
 			break;
+		
+		if (!strcasecmp(type, "action")) {
+			fprintf(stderr,
+				"Error: Resource type '%s' is reserved",
+				type);
+			free(type);
+			break;
+		}
 
 		rr = malloc(sizeof(*rr));
 		if (!rr)
--- cluster/rgmanager/src/daemons/rg_thread.c	2006/11/03 16:26:18	1.7.2.7
+++ cluster/rgmanager/src/daemons/rg_thread.c	2006/12/13 18:19:56	1.7.2.8
@@ -145,7 +145,7 @@
 		list_remove(list, curr);
 		dprintf("Removed request %d\n", curr->rr_request);
 		if (curr->rr_resp_fd != -1) {
-			send_response(RG_EABORT, curr);
+			send_response(RG_EABORT, NODE_ID_NONE, curr);
 		}
 		rq_free(curr);
 	}
@@ -412,9 +412,8 @@
 
 		if (ret != RG_NONE && rg_initialized() &&
 		    (req->rr_resp_fd >= 0)) {
-			send_response(error, req);
+			send_response(error, newowner, req);
 		}
-
 		
 		rq_free(req);
 	}
@@ -611,7 +610,7 @@
 		case RG_START:
 		case RG_ENABLE:
 			send_ret(response_fd, resgroup->rt_name, RG_EDEADLCK,
-				 request);
+				 request, NODE_ID_NONE);
 			break;
 		}
 		fprintf(stderr, "Failed to queue request: Would block\n");
--- cluster/rgmanager/src/daemons/reslist.c	2005/10/17 20:23:52	1.6.2.6
+++ cluster/rgmanager/src/daemons/reslist.c	2006/12/13 18:19:56	1.6.2.7
@@ -523,6 +523,83 @@
 }
 
 
+/* Copied from resrules.c -- _get_actions */
+void
+_get_actions_ccs(int ccsfd, char *base, resource_t *res)
+{
+	char xpath[256];
+	int idx = 0;
+	char *act, *ret;
+	int interval, timeout, depth;
+
+	do {
+		/* setting these to -1 prevents overwriting with 0 */
+		interval = -1;
+		depth = -1;
+		act = NULL;
+		timeout = -1;
+
+		snprintf(xpath, sizeof(xpath),
+			 "%s/action[%d]/@name", base, ++idx);
+
+#ifndef NO_CCS
+		if (ccs_get(ccsfd, xpath, &act) != 0)
+#else
+		if (conf_get(xpath, &act) != 0)
+#endif
+			break;
+
+		snprintf(xpath, sizeof(xpath),
+			 "%s/action[%d]/@timeout", base, idx);
+#ifndef NO_CCS
+		if (ccs_get(ccsfd, xpath, &ret) == 0 && ret) {
+#else
+		if (conf_get(xpath, &ret) == 0 && ret) {
+#endif
+			timeout = expand_time(ret);
+			if (timeout < 0)
+				timeout = 0;
+			free(ret);
+		}
+
+		snprintf(xpath, sizeof(xpath),
+			 "%s/action[%d]/@interval", base, idx);
+#ifndef NO_CCS
+		if (ccs_get(ccsfd, xpath, &ret) == 0 && ret) {
+#else
+		if (conf_get(xpath, &ret) == 0 && ret) {
+#endif
+			interval = expand_time(ret);
+			if (interval < 0)
+				interval = 0;
+			free(ret);
+		}
+
+		if (!strcmp(act, "status") || !strcmp(act, "monitor")) {
+			snprintf(xpath, sizeof(xpath),
+				 "%s/action[%d]/@depth", base, idx);
+#ifndef NO_CCS
+			if (ccs_get(ccsfd, xpath, &ret) == 0 && ret) {
+#else
+			if (conf_get(xpath, &ret) == 0 && ret) {
+#endif
+				depth = atoi(ret);
+				if (depth < 0)
+					depth = 0;
+				
+				/* */
+				if (ret[0] == '*')
+					depth = -1;
+				free(ret);
+			}
+		}
+
+		if (store_action(&res->r_actions, act, depth, timeout,
+				 interval) != 0)
+			free(act);
+	} while (1);
+}
+
 
 /**
    Try to load all the attributes in our rule set.  If none are found,
@@ -628,12 +705,12 @@
 	}
 
 	if (!found) {
-		//printf("No attributes found for %s\n", base);
 		destroy_resource(res);
 		return NULL;
 	}
 
 	res->r_actions = act_dup(rule->rr_actions);
+	_get_actions_ccs(ccsfd, base, res);
 
 	return res;
 }
@@ -660,7 +737,7 @@
 		for (resID = 1; ; resID++) {
 			snprintf(tok, sizeof(tok), RESOURCE_BASE "/%s[%d]",
 				 currule->rr_type, resID);
-
+			
 			newres = load_resource(ccsfd, currule, tok);
 			if (!newres)
 				break;
--- cluster/rgmanager/src/daemons/rg_state.c	2006/09/28 20:02:07	1.4.2.16
+++ cluster/rgmanager/src/daemons/rg_state.c	2006/12/13 18:19:56	1.4.2.17
@@ -111,123 +111,6 @@
 
 
 int
-clu_lock_verbose(char *resource, int dflt_flags, void **lockpp)
-{
-	int ret, timed_out = 0;
-	struct timeval start, now;
-	uint64_t nodeid, *p;
-	int flags;
-	int conv = 0, err;
-	int block = !(dflt_flags & CLK_NOWAIT);
-
-	/* Holder not supported for this call */
-	dflt_flags &= ~CLK_HOLDER;
-
-	flags = dflt_flags;
-
-	if (block) {
-		gettimeofday(&start, NULL);
-		start.tv_sec += 30;
-	}
-
-	/* Ripped from global.c in magma */
-	if (!(dflt_flags & CLK_CONVERT) &&
-	    (block || ((dflt_flags & CLK_EX) == 0))) {
-		/* Acquire NULL lock */
-		ret = clu_lock(resource, CLK_NULL, lockpp);
-		err = errno;
-		if (ret == 0) {
-			if ((flags & CLK_EX) == 0) {
-				/* User only wanted a NULL lock... */
-				return 0;
-			}
-			/*
-			   Ok, NULL lock was taken, rest of blocking
-			   call should be done using lock conversions.
-			 */
-			flags |= CLK_CONVERT;
-			conv = 1;
-		} else {
-			switch(err) {
-			case EINVAL:
-				/* Oops, null locks don't work on this
-				   plugin; use normal spam mode */
-				break;
-			default:
-				errno = err;
-				return -1;
-			}
-		}
-	}
-
-	while (1) {
-		if (block) {
-			gettimeofday(&now, NULL);
-
-			if ((now.tv_sec > start.tv_sec) ||
-			    ((now.tv_sec == start.tv_sec) &&
-	 		     (now.tv_usec >= start.tv_usec))) {
-
-				gettimeofday(&start, NULL);
-				start.tv_sec += 30;
-
-				timed_out = 1;
-				flags |= CLK_HOLDER;
-			}
-		}
-
-		*lockpp = NULL;
-
-		/* Take the lock (convert if possible). */
-		ret = clu_lock(resource, flags | CLK_NOWAIT |
-			       ((conv && !timed_out) ? CLK_CONVERT : 0),
-			       lockpp);
-		err = errno;
-
-		if ((ret != 0) && (err == EAGAIN) && block) {
-			if (timed_out) {
-				p = (uint64_t *)*lockpp;
-				if (p) {
-					nodeid = *p;
-					clulog(LOG_WARNING, "Node ID:%08x%08x"
-					       " stuck with lock %s\n",
-					       (uint32_t)(nodeid>>32&0xffffffff),
-					       (uint32_t)nodeid&0xffffffff,
-					       resource);
-					free(p);
-				} else {
-					clulog(LOG_WARNING, "Starving for lock"
-					       " %s\n", resource);
-				}
-				flags = dflt_flags;
-				timed_out = 0;
-			}
-			usleep(random()&32767<<1);
-			continue;
-
-		} else if (ret == 0) {
-			/* Success */
-			return 0;
-		}
-
-		break;
-	}
-
-	/* Fatal error.  If we took an automatic NL lock with the hopes of
-	   converting it, release the lock before returning */
-	if (conv == 1 && ret < 0) {
-		clu_unlock(resource, *lockpp);
-		*lockpp = NULL;
-	}
-
-	if (ret < 0)
-		errno = err;
-
-	return ret;
-}
-
-
-int
 #ifdef DEBUG
 _rg_lock(char *name, void **p)
 #else
@@ -237,7 +120,7 @@
 	char res[256];
 
 	snprintf(res, sizeof(res), "usrm::rg=\"%s\"", name);
-	return clu_lock_verbose(res, CLK_EX, p);
+	return clu_lock(res, CLK_EX, p);
 }
 
 
@@ -277,7 +160,7 @@
 
 
 void
-send_ret(int fd, char *name, int ret, int orig_request)
+send_ret(int fd, char *name, int ret, int orig_request, uint64_t newowner)
 {
 	SmMessageSt msg, *msgp = &msg;
 	if (fd < 0)
@@ -289,7 +172,11 @@
 	msgp->sm_data.d_action = orig_request;
 	strncpy(msgp->sm_data.d_svcName, name,
 		sizeof(msgp->sm_data.d_svcName));
-	msgp->sm_data.d_svcOwner = my_id(); /* XXX Broken */
+	if (newowner == NODE_ID_NONE) {
+		msgp->sm_data.d_svcOwner = my_id(); /* XXX Broken */
+	} else { 
+		msgp->sm_data.d_svcOwner = newowner;
+	}
 	msgp->sm_data.d_ret = ret;
 
 	swab_SmMessageSt(msgp);
@@ -301,7 +188,7 @@
 
 	
 void
-send_response(int ret, request_t *req)
+send_response(int ret, uint64_t newowner, request_t *req)
 {
 	SmMessageSt msg, *msgp = &msg;
 
@@ -314,7 +201,11 @@
 	msgp->sm_data.d_action = req->rr_orig_request;
 	strncpy(msgp->sm_data.d_svcName, req->rr_group,
 		sizeof(msgp->sm_data.d_svcName));
-	msgp->sm_data.d_svcOwner = my_id(); /* XXX Broken */
+	if (newowner == NODE_ID_NONE) {
+		msgp->sm_data.d_svcOwner = my_id(); /* XXX Broken */
+	} else { 
+		msgp->sm_data.d_svcOwner = newowner;
+	}
 	msgp->sm_data.d_ret = ret;
 
 	swab_SmMessageSt(msgp);
@@ -594,6 +485,7 @@
  *			1 = START service - return whatever it returns.
  *			2 = DO NOT start service, return 0
  *			3 = DO NOT start service, return RG_EAGAIN
+ *                      4 = DO NOT start servuce, return RG_ERUN
  */
 int
 svc_advise_start(rg_state_t *svcStatus, char *svcName, int req)
@@ -618,7 +510,7 @@
 			clulog(LOG_DEBUG,
 			       "RG %s is already running locally\n", svcName);
 			 */
-			ret = 2;
+			ret = 4;
 			break;
 		}
 
@@ -630,7 +522,7 @@
 			       svcName,
 			       memb_id_to_name(membership,svcStatus->rs_owner));
 			 */
-			ret = 2;
+			ret = 4;
 			break;
 		}
 
@@ -688,6 +580,7 @@
 		break;
 
 	case RG_STATE_STOPPED:
+	case RG_STATE_ERROR:
 		/* Don't actually enable if the RG is locked! */
 		if (rg_locked()) {
 			ret = 3;
@@ -719,7 +612,6 @@
 		       svcName);
 		break;
 
-	case RG_STATE_ERROR:
 	default:
 		clulog(LOG_ERR,
 		       "#44: Cannot start RG %s: Invalid State %d\n",
@@ -771,6 +663,9 @@
 	case 3:
 		rg_unlock(svcName, lockp);
 		return RG_EAGAIN;
+	case 4:
+		rg_unlock(svcName, lockp);
+		return RG_ERUN;
 	default:
 		break;
 	}
@@ -1404,8 +1299,8 @@
 	/* 
 	   If services are locked, return the error 
 	  */
-	if (ret == RG_EAGAIN)
-		return RG_EAGAIN;
+	if (ret == RG_EAGAIN || ret == RG_ERUN)
+		return ret;
 
 	/*
 	 * If we succeeded, then we're done.
--- cluster/rgmanager/src/daemons/main.c	2006/10/05 17:52:27	1.9.2.20
+++ cluster/rgmanager/src/daemons/main.c	2006/12/13 18:19:56	1.9.2.21
@@ -426,6 +426,7 @@
 		    	    sizeof (SmMessageSt))
 				clulog(LOG_ERR, "#40: Error replying to "
 				       "action request.\n");
+			break;
 		}
 
 		/* Queue request */
--- cluster/rgmanager/init.d/rgmanager	2006/09/07 18:39:45	1.3.2.3
+++ cluster/rgmanager/init.d/rgmanager	2006/12/13 18:19:56	1.3.2.4
@@ -96,10 +96,14 @@
 		[ -z "$RGMGR_OPTS" ] && RGMGR_OPTS="-t 30"
 		echo -n $"Starting $ID: "
 		daemon $RGMGRD $RGMGR_OPTS
+		rv=$?
 		echo
 
 		# To be consistent...
-		touch /var/lock/subsys/rgmanager
+		if [ $ret -eq 0 ]; then
+			touch /var/lock/subsys/rgmanager
+		fi
+		exit $rv
 		;;
 
 	restart)
--- cluster/rgmanager/src/clulib/vft.c	2006/05/12 21:28:31	1.7.2.6
+++ cluster/rgmanager/src/clulib/vft.c	2006/12/13 18:19:56	1.7.2.7
@@ -44,8 +44,6 @@
 #include <signals.h>
 
 
-int clu_lock_verbose(char *lockname, int flags, void **lockpp);
-
 static int vf_lfds[2];
 static int vf_lfd = 0;
 static key_node_t *key_list = NULL;	/** List of key nodes. */
@@ -1187,7 +1185,7 @@
 	pthread_mutex_lock(&vf_mutex);
 	/* Obtain cluster lock on it. */
 	snprintf(lock_name, sizeof(lock_name), "usrm::vf");
-	l = clu_lock_verbose(lock_name, CLK_EX, &lockp);
+	l = clu_lock(lock_name, CLK_EX, &lockp);
 	if (l < 0) {
 		clu_unlock(lock_name, lockp);
 		pthread_mutex_unlock(&vf_mutex);
@@ -1524,7 +1522,7 @@
 	/* Obtain cluster lock on it. */
 	pthread_mutex_lock(&vf_mutex);
 	snprintf(lock_name, sizeof(lock_name), "usrm::vf");
-	l = clu_lock_verbose(lock_name, CLK_EX, &lockp);
+	l = clu_lock(lock_name, CLK_EX, &lockp);
 	if (l < 0) {
 		clu_unlock(lock_name, lockp);
 		pthread_mutex_unlock(&vf_mutex);
--- cluster/rgmanager/src/utils/clusvcadm.c	2006/05/12 21:28:31	1.2.2.6
+++ cluster/rgmanager/src/utils/clusvcadm.c	2006/12/13 18:19:56	1.2.2.7
@@ -153,7 +153,7 @@
 printf("Resource Group Control Commands:\n");
 printf("       %s -v                     Display version and exit\n",name);
 printf("       %s -d <group>             Disable <group>\n", name);
-printf("       %s -e <group>             Enable <group>\n",
+printf("       %s -e <group>             Enable <group> on the local node\n",
        name);
 printf("       %s -e <group> -m <member> Enable <group>"
        " on <member>\n", name);
@@ -317,12 +317,31 @@
 		fprintf(stderr, "Error receiving reply!\n");
 		return 1;
 	}
-
+	
 	/* Decode */
 	swab_SmMessageSt(&msg);
 	switch (msg.sm_data.d_ret) {
 	case SUCCESS:
 		printf("success\n");
+
+		/* Non-start/relo request: done */
+		if (action != RG_RELOCATE && action != RG_ENABLE)
+			break;
+
+	    	if (svctarget != NODE_ID_NONE &&
+		    msg.sm_data.d_svcOwner != svctarget) {
+			/* Service running somewhere besides where requested */
+	    		printf("Warning: Service %s is running on %s "
+	    			"instead of %s\n", svcname,
+	    			memb_id_to_name(membership,
+				       msg.sm_data.d_svcOwner),
+				memb_id_to_name(membership, svctarget));
+			break;
+	    	}
+
+		/* No node specified or service running where requested */
+	    	printf("Service %s is now running on %s\n", svcname, 
+			memb_id_to_name(membership, msg.sm_data.d_svcOwner));
 		break;
 	case RG_EFAIL:
 		printf("failed\n");
@@ -339,6 +358,10 @@
 	case RG_EAGAIN:
 		printf("failed: Try again (resource groups locked)\n");
 		break;
+	case RG_ERUN:
+		printf("failed: Service is already running\n");
+		return 0;
+		break;
 	default:
 		printf("failed: unknown reason %d\n", msg.sm_data.d_ret);
 		break;
--- cluster/rgmanager/src/utils/clustat.c	2006/09/07 18:39:45	1.5.2.14
+++ cluster/rgmanager/src/utils/clustat.c	2006/12/13 18:19:56	1.5.2.15
@@ -358,14 +358,14 @@
 }
 
 
-void
+int
 txt_rg_states(rg_state_list_t *rgl, cluster_member_list_t *members, 
 	      char *svcname, int flags)
 {
-	int x;
+	int x, ret = 0;
 
 	if (!rgl || !members)
-		return;
+		return -1;
 
 	if (!(flags & RG_VERBOSE)) {
 		printf("  %-20.20s %-30.30s %-14.14s\n",
@@ -382,18 +382,31 @@
 		    strcmp(rgl->rgl_states[x].rs_name, svcname))
 			continue;
 		txt_rg_state(&rgl->rgl_states[x], members, flags);
+		if (svcname) {
+			switch (rgl->rgl_states[x].rs_state) {
+			case RG_STATE_STARTING:
+			case RG_STATE_STARTED:
+			case RG_STATE_STOPPING:
+				break;
+			default:
+				ret = rgl->rgl_states[x].rs_state;
+			}
+		}
 	}
+	
+	return ret;
 }
 
 
-void
+int
 xml_rg_states(rg_state_list_t *rgl, cluster_member_list_t *members,
 	      char *svcname)
 {
 	int x;
+	int ret = 0;
 
 	if (!rgl || !members)
-		return;
+		return -1;
 
 	printf("  <groups>\n");
 
@@ -401,14 +414,23 @@
 		if (svcname &&
 		    strcmp(rgl->rgl_states[x].rs_name, svcname))
 			continue;
-		xml_rg_state(&rgl->rgl_states[x], members, 0);
+		if (svcname) {
+			switch (rgl->rgl_states[x].rs_state) {
+			case RG_STATE_STARTING:
+			case RG_STATE_STARTED:
+			case RG_STATE_STOPPING:
+				break;
+			default:
+				ret = rgl->rgl_states[x].rs_state;
+			}
+		}
 	}
 
 	printf("  </groups>\n");
+	return ret;
 }
 
 
-
 void
 txt_quorum_state(int qs)
 {
@@ -481,14 +503,14 @@
 }
 
 
-void
+int 
 txt_member_states(cluster_member_list_t *membership, char *name)
 {
-	int x;
+	int x, ret = 0;
 
 	if (!membership) {
 		printf("Membership information not available\n");
-		return;
+		return -1;
 	}
 
 	printf("  %-40.40s %s\n", "Member Name", "Status");
@@ -498,20 +520,22 @@
 		if (name && strcmp(membership->cml_members[x].cm_name, name))
 			continue;
 		txt_member_state(&membership->cml_members[x]);
+		ret = !(membership->cml_members[x].cm_state & FLAG_UP);
 	}
 
 	printf("\n");
+	return ret;
 }
 
 
-void
+int 
 xml_member_states(cluster_member_list_t *membership, char *name)
 {
-	int x;
+	int x, ret = 0;
 
 	if (!membership) {
 		printf("  <nodes/>\n");
-		return;
+		return -1;
 	}
 
 	printf("  <nodes>\n");
@@ -519,16 +543,22 @@
 		if (name && strcmp(membership->cml_members[x].cm_name, name))
 			continue;
 		xml_member_state(&membership->cml_members[x]);
+		if (name)
+			ret = !(membership->cml_members[x].cm_state & FLAG_UP);
 	}
 	printf("  </nodes>\n");
+	
+	return ret;
 }
 
 
-void
+int 
 txt_cluster_status(int qs, cluster_member_list_t *membership,
 		   rg_state_list_t *rgs, char *name, char *svcname, 
 		   int flags)
 {
+	int ret;
+	
 	if (!svcname && !name) {
 		txt_quorum_state(qs);
 		if (!membership || !(qs & QF_GROUPMEMBER)) {
@@ -538,50 +568,43 @@
 	}
 
 	if (!svcname || (name && svcname))
-		txt_member_states(membership, name);
+		ret = txt_member_states(membership, name);
+	if (name && !svcname)
+		return ret;
 	if (!name || (name && svcname))
-		txt_rg_states(rgs, membership, svcname, flags);
+		ret = txt_rg_states(rgs, membership, svcname, flags);
+	return ret;
 }
 
 
-void
+int
 xml_cluster_status(int qs, cluster_member_list_t *membership,
 		   rg_state_list_t *rgs, char *name, char *svcname,
 		   int flags)
 {
+	int ret1 = 0, ret2 = -1;
+	
 	printf("<?xml version=\"1.0\"?>\n");
 	printf("<clustat version=\"4.1.1\">\n");
 
 	if (!svcname && !name)
 		xml_quorum_state(qs);
 	if (!svcname || (name && svcname)) 
-		xml_member_states(membership, name);
+		ret1 = xml_member_states(membership, name);
+	
 	if (rgs &&
 	    (!name || (name && svcname)))
-		xml_rg_states(rgs, membership, svcname);
+		ret2 = xml_rg_states(rgs, membership, svcname);
 	printf("</clustat>\n");
+	
+	if (name && ret1)
+		return ret1;
+	if (svcname && ret2)
+		return ret2;
+	return 0;
 }
 
 
-void
-dump_node(cluster_member_t *node)
-{
-	printf("Node %s state %02x\n", node->cm_name, node->cm_state);
-}
-
-
-void 
-dump_nodes(cluster_member_list_t *nodes)
-{
-	int x;
-
-	for (x=0; x<nodes->cml_count; x++) {
-		dump_node(&nodes->cml_members[x]);
-	}
-}
-
-
-
 cluster_member_list_t *
 build_member_list(uint64_t *lid)
 {
@@ -778,11 +801,13 @@
 		}
 
 		if (xml)
-			xml_cluster_status(qs, membership, rgs, member_name,
-					   rg_name,flags);
+			ret = xml_cluster_status(qs, membership, rgs,
+						 member_name, rg_name,
+						 flags);
 		else
-			txt_cluster_status(qs, membership, rgs, member_name,
-					   rg_name,flags);
+			ret = txt_cluster_status(qs, membership, rgs,
+						 member_name, rg_name,
+						 flags);
 
 		if (membership)
 			cml_free(membership);
--- cluster/rgmanager/src/resources/ip.sh	2006/05/16 20:03:04	1.5.2.15
+++ cluster/rgmanager/src/resources/ip.sh	2006/12/13 18:19:57	1.5.2.16
@@ -884,6 +884,9 @@
 		exit 0
 	fi
 	ip_op ${OCF_RESKEY_family} add ${OCF_RESKEY_address}
+	if [ $? -ne 0 ]; then
+		exit $OCF_ERR_GENERIC
+	fi
 
 	if [ $NFS_TRICKS -eq 0 ]; then
 		if [ "$OCF_RESKEY_nfslock" = "yes" ] || \
--- cluster/rgmanager/include/resgroup.h	2006/05/12 21:28:30	1.3.2.7
+++ cluster/rgmanager/include/resgroup.h	2006/12/13 18:19:57	1.3.2.8
@@ -121,8 +121,8 @@
 int rt_enqueue_request(const char *resgroupname, int request, int response_fd,
        		       int max, uint64_t target, int arg0, int arg1);
 
-void send_response(int ret, request_t *req);
-void send_ret(int fd, char *name, int ret, int req);
+void send_response(int ret, uint64_t owner, request_t *req);
+void send_ret(int fd, char *name, int ret, int req, uint64_t newowner);
 
 /* do this op on all resource groups.  The handler for the request 
    will sort out whether or not it's a valid request given the state */
@@ -156,12 +156,13 @@
 cluster_member_list_t *member_list(void);
 uint64_t my_id(void);
 
-#define RG_EAGAIN	-6
-#define RG_EDEADLCK	-5
-#define RG_ENOSERVICE	-4
-#define RG_EFORWARD	-3
-#define RG_EABORT	-2
-#define RG_EFAIL	-1
+#define RG_ERUN		-7 /* Service is running already */
+#define RG_EAGAIN	-6 /* Try again */
+#define RG_EDEADLCK	-5 /* Operation would cause deadlock */
+#define RG_ENOSERVICE	-4 /* Service does not exist */
+#define RG_EFORWARD	-3 /* Ask current service owner to do this, please */
+#define RG_EABORT	-2 /* Request cancelled */
+#define RG_EFAIL	-1 /* Generic error */
 #define RG_ESUCCESS	0
 
 


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]