Make ldmsd continue startup on producer's hostname resolution failure

ldms/man/ldmsd_controller.man

@@ -338,7 +338,10 @@ data our ldmsd holds for the prdcr.
 .br
 The recv rate (bytes/sec) limit for this connection. The default is -1
 (unlimited).
-
+.TP
+.BI [cache_ip " cache_ip"]
+.br
+Controls how \fBldmsd\fR handles hostname resolution for producer IP addresses. When set to \fBtrue\fR (default), \fBldmsd\fR resolves the hostname once during \fBprdcr_add\fR and caches the result. If the initial resolution fails and the producer is started (via \fBprdcr_start\R or \fBprdcr_start_regex\fR), \fBldmsd\fR will retry resolution at connection time and each resonnection attempt until successful. When set to \fBfalse\fR, \fBldmsd\fR performs hostname resolution at \fBprdcr_add\fR time and repeats the resolution at every connection and reconnection attempt if the producer is started.
 .RE
 
 .SS Delete a producer from the aggregator

ldms/python/ldmsd/ldmsd_communicator.py

@@ -2189,11 +2189,13 @@ def _prdcr_add_attr_prep(self, **kwargs):
             attrs.append(LDMSD_Req_Attr(attr_id=LDMSD_Req_Attr.QUOTA, value=str(int(kwargs['quota']))))
         if 'rx_rate' in kwargs.keys() and kwargs['rx_rate']:
             attrs.append(LDMSD_Req_Attr(attr_id=LDMSD_Req_Attr.RX_RATE, value=str(int(kwargs['rx_rate']))))
+        if 'cache_ip' in kwargs.keys() and kwargs['cache_ip']:
+            attrs.append(LDMSD_Req_Attr(attr_id = LDMSD_Req_Attr.IP, value = str(kwargs['cache_ip'])))
 
         return attrs
 
     def prdcr_add(self, name, ptype, xprt, host, port, reconnect, auth=None, perm=None,
-                  rail=None, quota=None, rx_rate=None):
+                  rail=None, quota=None, rx_rate=None, cache_ip=None):
         """
         Add a producer. A producer is a peer to the LDMSD being configured.
         Once started, the LDSMD will attempt to connect to this peer
@@ -2211,6 +2213,7 @@ def prdcr_add(self, name, ptype, xprt, host, port, reconnect, auth=None, perm=No
         - The reconnect interval in microseconds
 
         Keyword Parameters:
+        auth - The authentication domain
         perm - The configuration client permission required to
                modify the producer configuration. Default is None.
         rail - The number of endpoints in a rail. The default is 1.
@@ -2219,6 +2222,8 @@ def prdcr_add(self, name, ptype, xprt, host, port, reconnect, auth=None, perm=No
                   ('--quota' ldmsd option).
         rx_rate - The recv rate (bytes/second) limit for this connection. The
                   default is -1 (unlimited).
+        cache_ip - True: Cache hostname after first successfull resolution;
+                   False: Resolve hostname on every connection
 
         Returns:
         A tuple of status, data
@@ -2227,7 +2232,8 @@ def prdcr_add(self, name, ptype, xprt, host, port, reconnect, auth=None, perm=No
         """
         args_d = {'name': name, 'ptype': ptype, 'xprt': xprt, 'host': host, 'port': port,
                   'reconnect': reconnect, 'auth': auth, 'perm': perm,
-                  'rail': rail, 'quota': quota, 'rx_rate': rx_rate}
+                  'rail': rail, 'quota': quota, 'rx_rate': rx_rate,
+                  'cache_ip' : cache_ip}
         attrs = self._prdcr_add_attr_prep(**args_d)
         req = LDMSD_Request( command_id = LDMSD_Request.PRDCR_ADD, attrs = attrs)
         try:

ldms/python/ldmsd/ldmsd_controller

@@ -381,6 +381,8 @@ class LdmsdCmdParser(cmd.Cmd):
                    holds for the prdcr.
         [rx_rate=] The recv rate (bytes/sec) limit for this connection. The
                    default is -1 (unlimited).
+        [cache_ip=]   True to cache the IP address after first successful resolution (default).
+                      False to resolve the hostname at prdcr_add and at every connection attempt.
         """
         arg = self.handle_args('prdcr_add', arg)
         if arg is None:
@@ -404,7 +406,8 @@ class LdmsdCmdParser(cmd.Cmd):
                                           arg['perm'],
                                           arg['rail'],
                                           arg['quota'],
-                                          arg['rx_rate'])
+                                          arg['rx_rate'],
+                                          arg['cache_ip'])
             if rc:
                 print(f'Error adding prdcr {arg["name"]}: {msg}')
 

ldms/src/ldmsd/ldmsd.h

@@ -205,6 +205,11 @@ typedef struct ldmsd_prdcr_stream_s {
 typedef struct ldmsd_prdcr {
 	struct ldmsd_cfgobj obj;
 
+	/* Controls hostname resolution caching behavior (user configurable)
+	 * 1 = (default) Cache hostname after first successfull resolution
+	 * 0 = Resolve hostname on every connection
+	 */
+	uint8_t cache_ip;
 	struct sockaddr_storage ss;	/* Host address */
 	socklen_t ss_len;
 	char *host_name;	/* Host name */
@@ -1157,7 +1162,7 @@ ldmsd_prdcr_new_with_auth(const char *name, const char *xprt_name,
 		enum ldmsd_prdcr_type type,
 		int conn_intrvl_us,
 		const char *auth, uid_t uid, gid_t gid, int perm, int rail,
-		int64_t quota, int64_t rx_rate);
+		int64_t quota, int64_t rx_rate, int cache_ip);
 int ldmsd_prdcr_del(const char *prdcr_name, ldmsd_sec_ctxt_t ctxt);
 ldmsd_prdcr_t ldmsd_prdcr_first();
 ldmsd_prdcr_t ldmsd_prdcr_next(struct ldmsd_prdcr *prdcr);

ldms/src/ldmsd/ldmsd_failover.c

@@ -463,6 +463,12 @@ int __failover_send_prdcr(ldmsd_failover_t f, ldms_t x, ldmsd_prdcr_t p)
 	if (rc)
 		goto cleanup;
 
+	/* Cache_ip */
+	snprintf(buff, sizeof(buff), "%d", p->cache_ip);
+	rc = ldmsd_req_cmd_attr_append_str(rcmd, LDMSD_ATTR_IP, buff);
+	if (rc)
+		goto cleanup;
+
 	/* Terminate the message */
 	rc = ldmsd_req_cmd_attr_term(rcmd);
 	if (rc)
@@ -2050,6 +2056,7 @@ int failover_cfgprdcr_handler(ldmsd_req_ctxt_t req)
 	char *rail_s = __req_attr_gets(req, LDMSD_ATTR_RAIL);
 	char *quota_s = __req_attr_gets(req, LDMSD_ATTR_QUOTA);
 	char *rx_rate_s = __req_attr_gets(req, LDMSD_ATTR_RX_RATE);
+	char *cache_ip = __req_attr_gets(req, LDMSD_ATTR_IP);
 
 	uid_t _uid;
 	gid_t _gid;
@@ -2116,7 +2123,7 @@ int failover_cfgprdcr_handler(ldmsd_req_ctxt_t req)
 
 	p = ldmsd_prdcr_new_with_auth(name, xprt, host, atoi(port), ptype,
 			atoi(interval), auth, _uid, _gid, _perm, rail, quota,
-			rx_rate);
+			rx_rate, atoi(cache_ip));
 	if (!p) {
 		rc = errno;
 		str_rbn_free(srbn);

ldms/src/ldmsd/ldmsd_prdcr.c

@@ -854,6 +854,17 @@ static void prdcr_connect(ldmsd_prdcr_t prdcr)
 {
 	int ret;
 
+	if ((0 == prdcr->ss.ss_family) || (!prdcr->cache_ip)) {
+		if (prdcr_resolve(prdcr->host_name, prdcr->port_no,
+					&prdcr->ss, &prdcr->ss_len)) {
+			ovis_log(prdcr_log, OVIS_LERROR, "Producer '%s' connection failed. " \
+						"Hostname '%s:%u' not resolved.\n",
+						prdcr->obj.name, prdcr->host_name,
+						(unsigned) prdcr->port_no);
+			return;
+		}
+	}
+
 	switch (prdcr->type) {
 	case LDMSD_PRDCR_TYPE_ACTIVE:
 	case LDMSD_PRDCR_TYPE_BRIDGE:
@@ -995,7 +1006,7 @@ ldmsd_prdcr_new_with_auth(const char *name, const char *xprt_name,
 		const char *host_name, const unsigned short port_no,
 		enum ldmsd_prdcr_type type, int conn_intrvl_us,
 		const char *auth, uid_t uid, gid_t gid, int perm, int rail,
-		int64_t quota, int64_t rx_rate)
+		int64_t quota, int64_t rx_rate, int cache_ip)
 {
 	extern struct rbt *cfgobj_trees[];
 	struct ldmsd_prdcr *prdcr;
@@ -1014,6 +1025,7 @@ ldmsd_prdcr_new_with_auth(const char *name, const char *xprt_name,
 	prdcr->conn_intrvl_us = conn_intrvl_us;
 	prdcr->port_no = port_no;
 	prdcr->conn_state = LDMSD_PRDCR_STATE_STOPPED;
+	prdcr->cache_ip = cache_ip;
 	rbt_init(&prdcr->set_tree, set_cmp);
 	rbt_init(&prdcr->hint_set_tree, ldmsd_updtr_schedule_cmp);
 	prdcr->rail = rail;
@@ -1032,11 +1044,9 @@ ldmsd_prdcr_new_with_auth(const char *name, const char *xprt_name,
 	}
 
 	prdcr->ss_len = sizeof(prdcr->ss);
-	if (prdcr_resolve(host_name, port_no, &prdcr->ss, &prdcr->ss_len)) {
-		errno = EAFNOSUPPORT;
-		ovis_log(prdcr_log, OVIS_LERROR, "ldmsd_prdcr_new: %s:%u not resolved.\n",
-			host_name,(unsigned) port_no);
-		goto out;
+	if (prdcr_resolve(prdcr->host_name, prdcr->port_no, &prdcr->ss, &prdcr->ss_len)) {
+		ovis_log(config_log, OVIS_LWARN, "Producer '%s': %s:%u not resolved.\n",
+			prdcr->obj.name, prdcr->host_name,(unsigned) prdcr->port_no);
 	}
 
 	if (!auth)
@@ -1076,7 +1086,8 @@ ldmsd_prdcr_new(const char *name, const char *xprt_name,
 {
 	return ldmsd_prdcr_new_with_auth(name, xprt_name, host_name,
 			port_no, type, conn_intrvl_us,
-			DEFAULT_AUTH, getuid(), getgid(), 0777, rail, quota, rx_rate);
+			DEFAULT_AUTH, getuid(), getgid(), 0777, rail, quota,
+			rx_rate, 1);
 }
 
 extern struct rbt *cfgobj_trees[];

ldms/src/ldmsd/ldmsd_request.c

@@ -1647,7 +1647,9 @@ ldmsd_prdcr_t __prdcr_add_handler(ldmsd_req_ctxt_t reqc, char *verb, char *obj_n
 	int64_t quota = ldmsd_quota; /* use the global quota setting by default */
 	int64_t rx_rate = LDMS_UNLIMITED;
 	int rail = 1;
-	char *perm_s = NULL;
+	char *perm_s, *cache_ip_s;
+	int cache_ip = 1; /* Default is 1. */
+	perm_s = cache_ip_s = NULL;
 
 	name = host = xprt = type_s = port_s = interval_s = auth = rail_s = quota_s = NULL;
 
@@ -1773,9 +1775,17 @@ ldmsd_prdcr_t __prdcr_add_handler(ldmsd_req_ctxt_t reqc, char *verb, char *obj_n
 			goto out;
 		}
 	}
+
+	cache_ip_s = ldmsd_req_attr_str_value_get_by_id(reqc, LDMSD_ATTR_IP);
+	if (cache_ip_s) {
+		if (0 == strcasecmp(cache_ip_s, "false")) {
+			cache_ip = 0;
+		}
+	}
+
 	prdcr = ldmsd_prdcr_new_with_auth(name, xprt, host, port_no, type,
 					  interval_us, auth, uid, gid, perm,
-					  rail, quota, rx_rate);
+					  rail, quota, rx_rate, cache_ip);
 	if (!prdcr) {
 		if (errno == EEXIST)
 			goto eexist;
@@ -9988,7 +9998,7 @@ static int __process_advertisement(ldmsd_req_ctxt_t reqc, ldmsd_prdcr_listen_t l
 				NULL, uid, gid, 0770,
 				ldms_xprt_rail_eps(x),
 				ldms_xprt_rail_recv_quota_get(x),
-				ldms_xprt_rail_recv_rate_limit_get(x));
+				ldms_xprt_rail_recv_rate_limit_get(x), 1);
 		if (!prdcr) {
 			reqc->errcode = ENOMEM;
 			reqc->line_off = snprintf(reqc->line_buf, reqc->line_len,

ldms/src/ldmsd/ldmsd_request_util.c

@@ -183,6 +183,7 @@ struct req_str_id attr_str_id_table[] = {
 	{  "auto_interval",     LDMSD_ATTR_AUTO_INTERVAL  },
 	{  "auto_switch",       LDMSD_ATTR_AUTO_SWITCH  },
 	{  "base",              LDMSD_ATTR_BASE  },
+	{  "cache_ip",          LDMSD_ATTR_IP  },
 	{  "container",         LDMSD_ATTR_CONTAINER  },
 	{  "decomposition",     LDMSD_ATTR_DECOMP  },
 	{  "disable_start",     LDMSD_ATTR_AUTO_INTERVAL  },