orte: Add option for mixing of short and long hostnames

jjhursey · gpaulsen · commit 710912e846a5 · 2019-02-28T02:58:03.000-05:00
* Addresses Epic 103619
 * New MCA paramter orte_use_mixed_hostnames (default: 'false')
 * In some scenarios users expect the short hostname 'node01'
   to match the long hostname 'node01.mycluster.org'. There are cases
   in the runtime where the difference causes accounting errors (best case)
   or failure to launch (worst case).
   - This parameter causes us to strip off the FQDN from the hostname,
     leaving on the short name for comparison purposes.
   - This is similar to the orte_keep_fqdn_hostnames MCA parameter, except
     works as it's opposite. We did not want interprent the fqdn_hostnames
     parameter as being set to 'false' (default) as 'use only shortnames'
     since it is currenlty being interpreted differently for clusters
     where their might be significance in the short versus long hostnames.

(cherry picked from commit 48f3a00a645cb48653a79f2a5375763ce7b9e92e)

Conflicts:
	orte/mca/ras/base/ras_base_allocate.c
diff --git a/orte/mca/ras/base/ras_base_allocate.c b/orte/mca/ras/base/ras_base_allocate.c
@@ -14,6 +14,7 @@
  * Copyright (c) 2014-2018 Intel, Inc. All rights reserved.
  * Copyright (c) 2018      Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2016      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
diff --git a/orte/mca/rmaps/rank_file/rmaps_rank_file.c b/orte/mca/rmaps/rank_file/rmaps_rank_file.c
@@ -34,6 +34,9 @@
 #ifdef HAVE_UNISTD_H
 #include <unistd.h>
 #endif  /* HAVE_UNISTD_H */
+#if HAVE_ARPA_INET_H
+#include <arpa/inet.h>
+#endif
 #include <string.h>
 
 #include "opal/util/argv.h"
@@ -501,11 +504,16 @@ static int orte_rmaps_rank_file_parse(const char *rankfile)
                         }
                         opal_argv_free (argv);
 
-                        // Strip off the FQDN if present, ignore IP addresses
-                        if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(node_name) ) {
+                        // Strip off the FQDN if present
+                        if( orte_use_mixed_hostnames ) {
                             char *ptr;
-                            if (NULL != (ptr = strchr(node_name, '.'))) {
-                                *ptr = '\0';
+                            struct in_addr buf;
+                            /* if the nodename is an IP address, do not mess with it! */
+                            if (0 == inet_pton(AF_INET, node_name, &buf) &&
+                                0 == inet_pton(AF_INET6, node_name, &buf)) {
+                                if (NULL != (ptr = strchr(node_name, '.'))) {
+                                    *ptr = '\0';
+                                }
                             }
                         }
 
diff --git a/orte/mca/rmaps/seq/rmaps_seq.c b/orte/mca/rmaps/seq/rmaps_seq.c
@@ -192,11 +192,17 @@ static int orte_rmaps_seq_map(orte_job_t *jdata)
                 sq->cpuset = strdup(sep);
             }
 
-            // Strip off the FQDN if present, ignore IP addresses
-            if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(hstname) ) {
+            // Strip off the FQDN if present
+            if( orte_use_mixed_hostnames ) {
                 char *ptr;
-                if (NULL != (ptr = strchr(hstname, '.'))) {
-                    *ptr = '\0';
+                struct in_addr buf;
+
+                /* if the nodename is an IP address, do not mess with it! */
+                if (0 == inet_pton(AF_INET, hstname, &buf) &&
+                    0 == inet_pton(AF_INET6, hstname, &buf)) {
+                    if (NULL != (ptr = strchr(hstname, '.'))) {
+                        *ptr = '\0';
+                    }
                 }
             }
 
@@ -300,11 +306,16 @@ static int orte_rmaps_seq_map(orte_job_t *jdata)
                     sq->cpuset = strdup(sep);
                 }
 
-                // Strip off the FQDN if present, ignore IP addresses
-                if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(hstname) ) {
+                // Strip off the FQDN if present
+                if( orte_use_mixed_hostnames ) {
                     char *ptr;
-                    if (NULL != (ptr = strchr(hstname, '.'))) {
-                        (*ptr) = '\0';
+                    struct in_addr buf;
+                    /* if the nodename is an IP address, do not mess with it! */
+                    if (0 == inet_pton(AF_INET, hstname, &buf) &&
+                        0 == inet_pton(AF_INET6, hstname, &buf)) {
+                        if (NULL != (ptr = strchr(hstname, '.'))) {
+                            (*ptr) = '\0'; // JJH RETURN HERE TO DETERMINE WHY SEGV...
+                        }
                     }
                 }
 
diff --git a/orte/runtime/orte_globals.c b/orte/runtime/orte_globals.c
@@ -85,6 +85,7 @@ char *orte_oob_static_ports = NULL;
 bool orte_standalone_operation = false;
 bool orte_fwd_mpirun_port = true;
 
+bool orte_use_mixed_hostnames = false;
 bool orte_keep_fqdn_hostnames = false;
 bool orte_have_fqdn_allocation = false;
 bool orte_show_resolved_nodenames = false;
diff --git a/orte/runtime/orte_globals.h b/orte/runtime/orte_globals.h
@@ -470,6 +470,7 @@ ORTE_DECLSPEC extern bool orte_standalone_operation;
 ORTE_DECLSPEC extern bool orte_fwd_mpirun_port;
 
 /* nodename flags */
+ORTE_DECLSPEC extern bool orte_use_mixed_hostnames;
 ORTE_DECLSPEC extern bool orte_keep_fqdn_hostnames;
 ORTE_DECLSPEC extern bool orte_have_fqdn_allocation;
 ORTE_DECLSPEC extern bool orte_show_resolved_nodenames;
diff --git a/orte/runtime/orte_mca_params.c b/orte/runtime/orte_mca_params.c
@@ -422,6 +422,14 @@ int orte_register_params(void)
                                   OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_READONLY,
                                   &orte_keep_fqdn_hostnames);
 
+    /* whether or not to match short hostnames to FQDN hostnames */
+    orte_use_mixed_hostnames = false;
+    (void) mca_base_var_register ("orte", "orte", NULL, "use_mixed_hostnames",
+                                  "Whether or not to use mixed hostnames (short == FQDN) [default: no]",
+                                  MCA_BASE_VAR_TYPE_BOOL, NULL, 0, 0,
+                                  OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_READONLY,
+                                  &orte_use_mixed_hostnames);
+
     /* whether or not to retain aliases of hostnames */
     orte_retain_aliases = false;
     (void) mca_base_var_register ("orte", "orte", NULL, "retain_aliases",
diff --git a/orte/util/dash_host/dash_host.c b/orte/util/dash_host/dash_host.c
@@ -25,6 +25,9 @@
 
 #include <string.h>
 #include <ctype.h>
+#if HAVE_ARPA_INET_H
+#include <arpa/inet.h>
+#endif
 
 #include "orte/constants.h"
 #include "orte/types.h"
@@ -212,10 +215,17 @@ int orte_util_add_dash_host_nodes(opal_list_t *nodes,
             ndname = mini_map[i];
         }
 
-        // Strip off the FQDN if present, ignore IP addresses
-        if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(ndname) ) {
-            if (NULL != (ptr = strchr(ndname, '.'))) {
-                *ptr = '\0';
+ 
+        // Strip off the FQDN if present
+        if( orte_use_mixed_hostnames ) {
+            char *ptr;
+            struct in_addr buf;
+            /* if the nodename is an IP address, do not mess with it! */
+            if (0 == inet_pton(AF_INET, ndname, &buf) &&
+                0 == inet_pton(AF_INET6, ndname, &buf)) {
+                if (NULL != (ptr = strchr(ndname, '.'))) {
+                    *ptr = '\0';
+                }
             }
         }
         /* remove any modifier */
diff --git a/orte/util/hostfile/hostfile.c b/orte/util/hostfile/hostfile.c
@@ -28,6 +28,9 @@
 #ifdef HAVE_UNISTD_H
 #include <unistd.h>
 #endif
+#if HAVE_ARPA_INET_H
+#include <arpa/inet.h>
+#endif
 #include <errno.h>
 #include <string.h>
 #include <sys/stat.h>
@@ -166,11 +169,16 @@ static int hostfile_parse_line(int token, opal_list_t* updates,
         }
         opal_argv_free (argv);
 
-        // Strip off the FQDN if present, ignore IP addresses
-        if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(node_name) ) {
+        // Strip off the FQDN if present
+        if( orte_use_mixed_hostnames ) {
             char *ptr;
-            if (NULL != (ptr = strchr(node_name, '.'))) {
-                *ptr = '\0';
+            struct in_addr buf;
+            /* if the nodename is an IP address, do not mess with it! */
+            if (0 == inet_pton(AF_INET, node_name, &buf) &&
+                0 == inet_pton(AF_INET6, node_name, &buf)) {
+                if (NULL != (ptr = strchr(node_name, '.'))) {
+                    *ptr = '\0';
+                }
             }
         }
 
@@ -281,11 +289,16 @@ static int hostfile_parse_line(int token, opal_list_t* updates,
         }
         opal_argv_free (argv);
 
-        // Strip off the FQDN if present, ignore IP addresses
-        if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(node_name) ) {
+        // Strip off the FQDN if present
+        if( orte_use_mixed_hostnames ) {
             char *ptr;
-            if (NULL != (ptr = strchr(node_name, '.'))) {
-                *ptr = '\0';
+            struct in_addr buf;
+            /* if the nodename is an IP address, do not mess with it! */
+            if (0 == inet_pton(AF_INET, node_name, &buf) &&
+                0 == inet_pton(AF_INET6, node_name, &buf)) {
+                if (NULL != (ptr = strchr(node_name, '.'))) {
+                    *ptr = '\0';
+                }
             }
         }
 
diff --git a/orte/util/proc_info.c b/orte/util/proc_info.c
@@ -49,6 +49,7 @@
 
 /* provide a connection to a reqd variable */
 extern bool orte_keep_fqdn_hostnames;
+extern bool orte_use_mixed_hostnames;
 
 #define ORTE_NAME_INVALID {ORTE_JOBID_INVALID, ORTE_VPID_INVALID}
 
@@ -169,12 +170,18 @@ int orte_proc_info(void)
     /* add this to our list of aliases */
     opal_argv_append_nosize(&orte_process_info.aliases, hostname);
 
-    // Strip off the FQDN if present, ignore IP addresses
-    if( !orte_keep_fqdn_hostnames && !opal_net_isaddr(hostname) ) {
-        if (NULL != (ptr = strchr(hostname, '.'))) {
-            *ptr = '\0';
-            /* add this to our list of aliases */
-            opal_argv_append_nosize(&orte_process_info.aliases, hostname);
+    // These two options should be mutually exclusive
+    // Cannot specify orte_use_mixed_hostnames=true and orte_keep_fqdn_hostnames=true
+    if (!orte_keep_fqdn_hostnames || orte_use_mixed_hostnames) {
+        /* if the nodename is an IP address, do not mess with it! */
+        if (0 == inet_pton(AF_INET, hostname, &buf) &&
+            0 == inet_pton(AF_INET6, hostname, &buf)) {
+            /* not an IP address, so remove any domain info */
+            if (NULL != (ptr = strchr(hostname, '.'))) {
+                *ptr = '\0';
+                /* add this to our list of aliases */
+                opal_argv_append_nosize(&orte_process_info.aliases, hostname);
+            }
         }
     }
 

Original file line number	Diff line number	Diff line change
`@@ -14,6 +14,7 @@`
`14`	`14`	`* Copyright (c) 2014-2018 Intel, Inc. All rights reserved.`
`15`	`15`	`* Copyright (c) 2018 Research Organization for Information Science`
`16`	`16`	`* and Technology (RIST). All rights reserved.`
	`17`	`+ * Copyright (c) 2016 IBM Corporation. All rights reserved.`
`17`	`18`	`* $COPYRIGHT$`
`18`	`19`	`*`
`19`	`20`	`* Additional copyrights may follow`