Use peer's hostname and listen port for advertised producer's names

Currently, an advertised producer's name is the name of the advertiser
on the peer. This creates confusion. The patch constructs the producer
name as <hostname>:<port>, where hostname is the peer hostname and port
is the listening port. If there are multiple listening ports, ldmsd uses
the first listening port.

ldms/man/ldmsd_peer_daemon_advertisement.man

@@ -1,4 +1,5 @@
-a prdcr_listenv 2024" "v5" "LDMSD Peer Daemon Advertisement man page"
+\" Manpage for ldmsd_peer_daemon_advertisement
+.TH man 7 "12 December 2024" "v4" "LDMSD Peer Daemon Advertisement man page"
 
 .\""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""/.
 .SH NAME
@@ -7,7 +8,7 @@ ldmsd_peer_daemon_advertisement - Manual for LDMSD Peer Daemon Advertisement
 .\""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""/.
 .SH SYNOPSIS
 
-**Sampler side Commands**
+**Peer side Commands**
 
 .IP \fBadvertiser_add
 .RI "name=" NAME " xprt=" XPRT " host=" HOST " port=" PORT " reconnect=" RECONNECT
@@ -58,19 +59,19 @@ hostname to the aggregator. On the aggregator, admins specify a regular
 expression to be matched with the peer hostname or an IP range that the peer IP
 address falls in via the \fBprdcr_listen_add\fR command. The
 \fBprdcr_listen_start\fR command is used to tell the aggregator to
-automatically add producers corresponding to a peer daemon whose
-hostname matches the regular expression or whose IP address falls in the IP
-range. If neither a regular expression nor an IP range is given, the aggregator
-will create a producer upon receiving any advertisement messages.
+automatically add producers corresponding to a peer daemon whose hostname
+matches the regular expression or whose IP address falls in the IP range. If
+neither a regular expression nor an IP range is given, the aggregator will
+create a producer upon receiving any advertisement messages.
 
 The auto-generated producers are of the ‘advertised’ type. The producer name is
-the same as the name given at the \fBadvertiser_add\fR or
-\fBadvertiser_start\fR line in the peer daemon configuration file. LDMSD
-automatically starts the advertised producers, unless the 'disable_start'
-attribute is given on the \fBprdcr_listen_add\fR line.  The advertised
-producers need to be stopped manually by using the command \fBprdcr_stop\fR or
-\fBprdcr_stop_regex\fR. They can be restarted by using the command
-\fBprdcr_start\fR or \fBprdcr_start_regex\fR.
+\fB<host:port>\fR, where \fBhost\fR is the peer hostname, and \fBport\fR is the
+first listening port of the peer daemon. LDMSD automatically starts the
+advertised producers, unless the 'disable_start' attribute is given on the
+\fBprdcr_listen_add\fR line. The advertised producers need to be stopped
+manually by using the command \fBprdcr_stop\fR or \fBprdcr_stop_regex\fR. They
+can be restarted by using the command \fBprdcr_start\fR or
+\fBprdcr_start_regex\fR.
 
 The description for each command and its parameters are as follows.
 
@@ -79,7 +80,7 @@ The description for each command and its parameters are as follows.
 \fBadvertiser_add\fR adds a new advertisement. The parameters are:
 .RS
 .IP \fBname\fR=\fINAME
-String of the advertisement name. The aggregator uses the string as the producer name as well.
+Advertiser name
 .IP \fBhost\fR=\fIHOST
 Aggregator hostname
 .IP \fBxprt\fR=\fIXPRT
@@ -92,10 +93,12 @@ Reconnect interval
 The authentication domain to be used to connect to the aggregator
 .RE
 
-\fBadvertiser_start\fR starts an advertisement. If the advertiser does not exist, LDMSD will create the advertiser. In this case, the mandatory attributes for \fBadvertiser_add\fB must be given. The parameters are:
+\fBadvertiser_start\fR starts an advertisement. If the advertiser does not
+exist, LDMSD will create the advertiser. In this case, the mandatory attributes
+for \fBadvertiser_add\fB must be given. The parameters are:
 .RS
 .IP \fBname\fR=\fINAME
-The advertisement name to be started
+Name of the advertiser to be started
 .IP \fB[host\fR=\fIHOST\fB]
 Aggregator hostname
 .IP \fB[xprt\fR=\fIXPRT\fB]
@@ -111,19 +114,19 @@ The authentication domain to be used to connect to the aggregator
 \fBadvertiser_stop\fR stops an advertisement. The parameters are:
 .RS
 .IP \fBname\fR=\fINAME
-The advertisement name to be stopped
+Nmae of the advertiser to be stopped
 .RE
 
 \fBadvertiser_del\fR deletes an advertisement. The parameters are:
 .RS
 .IP \fBname\fR=\fINAME
-The advertisement name to be deleted
+Name of the advertiser to be deleted
 .RE
 
 \fBadvertiser_status reports the status of each advertisement. An optional parameter is:
 .RS
 .IP \fB[name\fR=\fINAME\fB]
-Advertisement name
+Advertiser name
 .RE
 
 .PP
@@ -166,20 +169,22 @@ Name of prdcr_listen to be deleted
 In this example, there are three LDMS daemons running on \fBnode-1\fR,
 \fBnode-2\fR, and \fBnode03\fR. LDMSD running on \fBnode-1\fR and \fBnode-2\fR
 are sampler daemons, namely \fBsamplerd-1\fR and \fBsamplerd-2\fR. The
-aggregator (\fBagg\fR) runs on \fBnode-3\fR. All LDMSD listen on port 411.
+aggregator (\fBagg11\fR) runs on \fBnode-3\fR. All LDMSD listen on port 411.
 
 The sampler daemons collect the \fBmeminfo\fR set, and they are configured to
 advertise themselves and connect to the aggregator using sock on host
-\fBnode-3\fR at port 411. The following are the configuration files of the
-\fBsamplerd-1\fR and \fBsamplerd-2\fR.
+\fBnode-3\fR at port 411. They will try to reconnect to the aggregator every 10
+seconds until the connection is established. Once the connection is
+established, they will send an advertisement to the aggregator. The following
+are the configuration files of the \fBsamplerd-1\fR and \fBsamplerd-2\fR.
 
 .EX
 .B
 > cat samplerd-1.conf
 .RS 4
 # Add and start an advertisement
-advertiser_add name=samplerd-1 xprt=sock host=node-3 port=411 reconnect=10s
-advertiser_start name=samplerd-1
+advertiser_add name=agg11 xprt=sock host=node-3 port=411 reconnect=10s
+advertiser_start name=agg11
 # Load, configure, and start the meminfo plugin
 load name=meminfo
 config name=meminfo producer=samplerd-1 instance=samplerd-1/meminfo
@@ -190,7 +195,7 @@ start name=meminfo interval=1s
 > cat samplerd-2.conf
 .RS 4
 # Add and start an advertisement using only the advertiser_start command
-advertiser_start name=samplerd-2 host=node-3 port=411 reconnect=10s
+advertiser_start name=agg11 host=node-3 port=411 reconnect=10s
 # Load, configure, and start the meminfo plugin
 load name=meminfo
 config name=meminfo producer=samplerd-2 instance=samplerd-2/meminfo
@@ -199,11 +204,7 @@ start name=meminfo interval=1s
 .EE
 
 The aggregator is configured to accept advertisements from the sampler daemons
-that the hostnames match the regular expressions \fBnode0[1-2]\fR. The
-auto-added producers will check for an establish connection with the samplers
-every 10 seconds if the connection becomes disconnected. An updater is added to
-update the sets of all producers on the aggregators every 10 seconds at the 100
-milliseconds offset.
+that the hostnames match the regular expressions \fBnode0[1-2]\fR.
 
 .EX
 .B
@@ -215,7 +216,7 @@ prdcr_listen_start name=computes
 # Add and start an updater
 updtr_add name=all_sets interval=1s offset=100ms
 updtr_prdcr_add name=all_sets regex=.*
-updtr_start name=all
+updtr_start name=all_sets
 .RE
 .EE
 
@@ -224,13 +225,13 @@ advertisement of a sampler daemon.
 
 .EX
 .B
-> ldmsd_controller -x sock -p 10001 -h node-1
+> ldmsd_controller -x sock -p 411 -h node-1
 Welcome to the LDMSD control processor
-sock:node-1:10001> advertiser_status
+sock:node-1:411> advertiser_status
 Name             Aggregator Host  Aggregator Port Transport    Reconnect (us)         State
 ---------------- ---------------- --------------- ------------ --------------- ------------
-samplerd-1                 node-3           10001         sock        10000000    CONNECTED
-sock:node-1:10001>
+agg11                      node-3             411         sock        10000000    CONNECTED
+sock:node-1:411>
 .EE
 
 Similarly, LDMSD provides the command \fBprdcr_listen_status\fR to report the
@@ -239,14 +240,14 @@ the list of auto-added producers corresponding to each prdcr_listen object.
 
 .EX
 .B
-> ldmsd_controller -x sock -p 10001 -h node-3
+> ldmsd_controller -x sock -p 411 -h node-3
 Welcome to the LDMSD control processor
-sock:node-3:10001> prdcr_listen_status
+sock:node-3:411> prdcr_listen_status
 Name                 State      Regex           IP Range
 -------------------- ---------- --------------- ------------------------------
 computes             running    node-[1-2]      -
-Producers: samplerd-1, samplerd-2
-sock:node-3:10001>
+Producers: node-1:411, node-2:411
+sock:node-3:411>
 .EE
 
 .SH SEE ALSO

ldms/src/ldmsd/ldmsd_prdcr.c

@@ -628,32 +628,54 @@ static int __advertise_resp_cb(ldmsd_req_cmd_t rcmd)
 	return 0;
 }
 
-static int __send_advertisement(ldmsd_prdcr_t prdcr)
+static void __send_advertisement(ldmsd_prdcr_t prdcr)
 {
 	int rc;
 	ldmsd_req_cmd_t rcmd;
+	ldmsd_listen_t l;
 	char my_hostname[HOST_NAME_MAX+1];
+	char lport[10];
 
 	rcmd = ldmsd_req_cmd_new(prdcr->xprt,
 				LDMSD_ADVERTISE_REQ, NULL,
 				__advertise_resp_cb, prdcr);
 	if (!rcmd) {
 		ldmsd_log(LDMSD_LCRITICAL, "Memory allocation failure.\n");
-		return ENOMEM;
+		goto out;
 	}
 
 	rc = ldmsd_req_cmd_attr_append_str(rcmd, LDMSD_ATTR_NAME, prdcr->obj.name);
-	if (rc)
-		goto out;
+	if (rc) {
+		ldmsd_log(LDMSD_LERROR, "Failed to construct an advertisement. " \
+								"Error %d\n", rc);
+		goto err;
+	}
 	rc = gethostname(my_hostname, HOST_NAME_MAX+1);
 	rc = ldmsd_req_cmd_attr_append_str(rcmd, LDMSD_ATTR_HOST, my_hostname);
-	if (rc)
-		goto out;
+	if (rc) {
+		ldmsd_log(LDMSD_LERROR, "Failed to construct an advertisement. " \
+								"Error %d\n", rc);
+		goto err;
+	}
+	l = (ldmsd_listen_t)ldmsd_cfgobj_first(LDMSD_CFGOBJ_LISTEN);
+	snprintf(lport, 10, "%d", l->port_no);
+	rc = ldmsd_req_cmd_attr_append_str(rcmd, LDMSD_ATTR_PORT, lport);
+	if (rc) {
+		ldmsd_log(LDMSD_LERROR, "Failed to cosntruct an advertisement. " \
+								"error %d\n", rc);
+		goto err;
+	}
 	rc = ldmsd_req_cmd_attr_term(rcmd);
-	if (rc)
-		goto out;
+	if (rc) {
+		ldmsd_log(LDMSD_LERROR, "Failed to send an advertisement. " \
+								"Error %d\n", rc);
+		goto err;
+	}
 out:
-	return rc;
+	return;
+err:
+	ldmsd_req_cmd_free(rcmd);
+	return;
 }
 
 static int __sampler_routine(ldms_t x, ldms_xprt_event_t e, ldmsd_prdcr_t prdcr)
@@ -665,7 +687,6 @@ static int __sampler_routine(ldms_t x, ldms_xprt_event_t e, ldmsd_prdcr_t prdcr)
 			prdcr->conn_state = LDMSD_PRDCR_STATE_CONNECTED;
 			if (prdcr->type == LDMSD_PRDCR_TYPE_ADVERTISER) {
 				__send_advertisement(prdcr);
-				/* TODO: handle the error */
 			}
 			break;
 		case LDMS_XPRT_EVENT_DISCONNECTED:

ldms/src/ldmsd/ldmsd_request.c

@@ -8930,10 +8930,11 @@ extern void prdcr_connect_cb(ldms_t x, ldms_xprt_event_t e, void *cb_arg);
 static int __process_advertisement(ldmsd_req_ctxt_t reqc, ldmsd_prdcr_listen_t lp, struct ldms_addr *rem_addr)
 {
 	int rc = 0;
-	char *name;
 	char *xprt_s;
 	char *hostname;
 	char *attr_name;
+	char *lport;
+	char name[NI_MAXHOST + NI_MAXSERV + 1];
 	ldmsd_prdcr_t prdcr;
 	ldmsd_prdcr_ref_t pl_pref, updtr_pref;
 	struct rbn *rbn;
@@ -8942,17 +8943,18 @@ static int __process_advertisement(ldmsd_req_ctxt_t reqc, ldmsd_prdcr_listen_t l
 	gid_t gid;
 	int is_start = 0;
 	struct ldms_xprt_event conn_ev;
-	name = xprt_s = hostname = NULL;
-
-	attr_name = "name";
-	name = ldmsd_req_attr_str_value_get_by_id(reqc, LDMSD_ATTR_NAME);
-	if (!name)
-		goto einval;
+	xprt_s = hostname = lport = NULL;
 
 	attr_name = "hostname";
 	hostname = ldmsd_req_attr_str_value_get_by_id(reqc, LDMSD_ATTR_HOST);
 	if (!hostname)
 		goto einval;
+	attr_name = "port";
+	lport = ldmsd_req_attr_str_value_get_by_id(reqc, LDMSD_ATTR_PORT);
+	if (!lport)
+		goto einval;
+
+	snprintf(name, 32, "%s:%s", hostname, lport);
 
 	xprt_s = (char *)ldms_xprt_type_name(reqc->xprt->ldms.ldms);
 
@@ -9150,7 +9152,6 @@ static int advertise_handler(ldmsd_req_ctxt_t reqc)
 						"The attribute 'hostname' is required.");
 		goto send_reply;
 	}
-
 	for (pl = (ldmsd_prdcr_listen_t)ldmsd_cfgobj_first(LDMSD_CFGOBJ_PRDCR_LISTEN);
 			pl; pl = (ldmsd_prdcr_listen_t)ldmsd_cfgobj_next(&pl->obj))
 	{