Begin handling the case of lost connections by having the OOB report it to the errmgr instead of the routed framework. Add an "app" component to t

he errmgr framework so that it can decide how to respond - which for now at least is just to check for lifeline and abort if so.

Add a new error constant to indicate that the error is "unrecoverable" so the oob can know it needs to abort.

This commit was SVN r23112.

orte/include/orte/constants.h

@@ -114,7 +114,8 @@ enum {
     ORTE_ERR_SYSTEM_WILL_BOOTSTRAP          = (ORTE_ERR_BASE - 31),
     ORTE_ERR_RELOCATE_LIMIT_EXCEEDED        = (ORTE_ERR_BASE - 32),
     ORTE_ERR_INVALID_NODE_RANK              = (ORTE_ERR_BASE - 33),
-    ORTE_ERR_INVALID_LOCAL_RANK             = (ORTE_ERR_BASE - 34)
+    ORTE_ERR_INVALID_LOCAL_RANK             = (ORTE_ERR_BASE - 34),
+    ORTE_ERR_UNRECOVERABLE                  = (ORTE_ERR_BASE - 35)
 };
 
 #define ORTE_ERR_MAX                      (ORTE_ERR_BASE - 100)

orte/mca/errmgr/app/.windows

@@ -0,0 +1,12 @@
+#
+# Copyright (c) 2008-2010 High Performance Computing Center Stuttgart, 
+#                         University of Stuttgart.  All rights reserved.
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+# Specific to this module
+mca_link_libraries=libopen-rte

orte/mca/errmgr/app/Makefile.am

@@ -0,0 +1,36 @@
+#
+# Copyright (c) 2010      Cisco Systems, Inc.  All rights reserved. 
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+EXTRA_DIST = .windows
+
+sources = \
+        errmgr_app.h \
+        errmgr_app_component.c \
+        errmgr_app.c
+
+# Make the output library in this directory, and name it either
+# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
+# (for static builds).
+
+if OMPI_BUILD_errmgr_app_DSO
+component_noinst =
+component_install = mca_errmgr_app.la
+else
+component_noinst = libmca_errmgr_app.la
+component_install =
+endif
+
+mcacomponentdir = $(pkglibdir)
+mcacomponent_LTLIBRARIES = $(component_install)
+mca_errmgr_app_la_SOURCES = $(sources)
+mca_errmgr_app_la_LDFLAGS = -module -avoid-version
+
+noinst_LTLIBRARIES = $(component_noinst)
+libmca_errmgr_app_la_SOURCES =$(sources)
+libmca_errmgr_app_la_LDFLAGS = -module -avoid-version

orte/mca/errmgr/app/configure.params

@@ -0,0 +1,22 @@
+# -*- shell-script -*-
+#
+# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
+#                         University Research and Technology
+#                         Corporation.  All rights reserved.
+# Copyright (c) 2004-2005 The University of Tennessee and The University
+#                         of Tennessee Research Foundation.  All rights
+#                         reserved.
+# Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
+#                         University of Stuttgart.  All rights reserved.
+# Copyright (c) 2004-2005 The Regents of the University of California.
+#                         All rights reserved.
+# Copyright (c) 2007      Los Alamos National Security, LLC.  All rights
+#                         reserved. 
+# $COPYRIGHT$
+# 
+# Additional copyrights may follow
+# 
+# $HEADER$
+#
+
+PARAM_CONFIG_FILES="Makefile"

orte/mca/errmgr/app/errmgr_app.c

@@ -0,0 +1,108 @@
+/*
+ * Copyright (c) 2009-2010 The Trustees of Indiana University.
+ *                         All rights reserved.
+ *
+ * Copyright (c) 2010      Cisco Systems, Inc.  All rights reserved. 
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+
+#include <sys/types.h>
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif  /* HAVE_UNISTD_H */
+#ifdef HAVE_STRING_H
+#include <string.h>
+#endif
+
+#include "opal/util/output.h"
+
+#include "orte/util/error_strings.h"
+#include "orte/util/name_fns.h"
+#include "orte/util/show_help.h"
+#include "orte/runtime/orte_globals.h"
+#include "orte/mca/routed/routed.h"
+
+#include "orte/mca/errmgr/base/base.h"
+#include "orte/mca/errmgr/base/errmgr_private.h"
+#include "errmgr_app.h"
+
+/*
+ * Module functions: Global
+ */
+static int init(void);
+static int finalize(void);
+
+static int update_state(orte_jobid_t job,
+                        orte_job_state_t jobstate,
+                        orte_process_name_t *proc_name,
+                        orte_proc_state_t state,
+                        orte_exit_code_t exit_code,
+                        orte_errmgr_stack_state_t *stack_state);
+
+/******************
+ * HNP module
+ ******************/
+orte_errmgr_base_module_t orte_errmgr_app_module = {
+    init,
+    finalize,
+    update_state,
+    NULL,
+    NULL,
+    NULL
+};
+
+/************************
+ * API Definitions
+ ************************/
+static int init(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static int finalize(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static int update_state(orte_jobid_t job,
+                        orte_job_state_t jobstate,
+                        orte_process_name_t *proc,
+                        orte_proc_state_t state,
+                        orte_exit_code_t exit_code,
+                        orte_errmgr_stack_state_t *stack_state)
+{
+    /* indicate that this is the end of the line */
+    *stack_state |= ORTE_ERRMGR_STACK_STATE_COMPLETE;
+    
+    OPAL_OUTPUT_VERBOSE((1, orte_errmgr_base.output,
+                         "%s errmgr:app: job %s reported state %s"
+                         " for proc %s state %s exit_code %d",
+                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                         ORTE_JOBID_PRINT(job),
+                         orte_job_state_to_str(jobstate),
+                         (NULL == proc) ? "NULL" : ORTE_NAME_PRINT(proc),
+                         orte_proc_state_to_str(state), exit_code));
+    
+    /*
+     * if orte is trying to shutdown, just let it
+     */
+    if (orte_finalizing) {
+        return ORTE_SUCCESS;
+    }
+
+    if (ORTE_PROC_STATE_COMM_FAILED == state) {
+        /* delete the route */
+        orte_routed.delete_route(proc);
+        /* see is this was a lifeline */
+        if (ORTE_SUCCESS != orte_routed.route_lost(proc)) {
+            return ORTE_ERR_UNRECOVERABLE;
+        }
+    }
+    return ORTE_SUCCESS;
+}

orte/mca/errmgr/app/errmgr_app.h

@@ -0,0 +1,35 @@
+/*
+ * Copyright (c) 2010      Cisco Systems, Inc. All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+/**
+ * @file
+ * 
+ */
+
+#ifndef MCA_ERRMGR_APP_EXPORT_H
+#define MCA_ERRMGR_APP_EXPORT_H
+
+#include "orte_config.h"
+
+#include "orte/mca/errmgr/errmgr.h"
+
+BEGIN_C_DECLS
+
+/*
+ * Local Component structures
+ */
+
+ORTE_MODULE_DECLSPEC extern orte_errmgr_base_component_t mca_errmgr_app_component;
+
+ORTE_DECLSPEC extern orte_errmgr_base_module_t orte_errmgr_app_module;
+
+END_C_DECLS
+
+#endif /* MCA_ERRMGR_APP_EXPORT_H */

orte/mca/errmgr/app/errmgr_app_component.c

@@ -0,0 +1,83 @@
+/*
+ * Copyright (c) 2010      Cisco Systems, Inc. All rights reserved.
+ *
+ * $COPYRIGHT$
+ * 
+ * Additional copyrights may follow
+ * 
+ * $HEADER$
+ */
+
+#include "orte_config.h"
+#include "opal/util/output.h"
+
+#include "orte/mca/errmgr/errmgr.h"
+#include "orte/mca/errmgr/base/base.h"
+#include "errmgr_app.h"
+
+/*
+ * Public string for version number
+ */
+const char *orte_errmgr_app_component_version_string = 
+    "ORTE ERRMGR app MCA component version " ORTE_VERSION;
+
+/*
+ * Local functionality
+ */
+static int errmgr_app_open(void);
+static int errmgr_app_close(void);
+static int errmgr_app_component_query(mca_base_module_t **module, int *priority);
+
+/*
+ * Instantiate the public struct with all of our public information
+ * and pointer to our public functions in it
+ */
+orte_errmgr_base_component_t mca_errmgr_app_component =
+{
+    /* Handle the general mca_component_t struct containing 
+     *  meta information about the component itapp
+     */
+    {
+        ORTE_ERRMGR_BASE_VERSION_3_0_0,
+        /* Component name and version */
+        "app",
+        ORTE_MAJOR_VERSION,
+        ORTE_MINOR_VERSION,
+        ORTE_RELEASE_VERSION,
+        
+        /* Component open and close functions */
+        errmgr_app_open,
+        errmgr_app_close,
+        errmgr_app_component_query
+    },
+    {
+        /* The component is checkpoint ready */
+        MCA_BASE_METADATA_PARAM_CHECKPOINT
+    }
+};
+
+static int errmgr_app_open(void) 
+{
+    return ORTE_SUCCESS;
+}
+
+static int errmgr_app_close(void)
+{
+    return ORTE_SUCCESS;
+}
+
+static int errmgr_app_component_query(mca_base_module_t **module, int *priority)
+{
+    if (ORTE_PROC_IS_APP) {
+        /* keep our priority low so that other modules are higher
+         * and will run before us
+         */
+        *priority = 10;
+        *module = (mca_base_module_t *)&orte_errmgr_app_module;
+        return ORTE_SUCCESS;        
+    }
+    
+    *priority = -1;
+    *module = NULL;
+    return ORTE_ERROR;
+}

orte/mca/errmgr/base/errmgr_base_fns.c

@@ -68,19 +68,13 @@ int orte_errmgr_base_update_state(orte_jobid_t job,
     int i;
     orte_errmgr_stack_state_t stack_state;
     orte_errmgr_base_module_t *module;
-
-    if( ORTE_PROC_IS_APP ) {
-        return ORTE_SUCCESS;
-    }
-
-    if( !orte_errmgr_base.shutting_down ) {
-        OPAL_OUTPUT_VERBOSE((10, orte_errmgr_base.output,
-                             "errmgr:base:update_state() %s) "
-                             "------- %s state updated for process %s",
-                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
-                             (NULL == name) ? "App. Process" : (name->jobid == ORTE_PROC_MY_HNP->jobid ? "Daemon" : "App. Process"),
-                             (NULL == name) ? "NULL" : ORTE_NAME_PRINT(name)));
-    }
+    
+    OPAL_OUTPUT_VERBOSE((10, orte_errmgr_base.output,
+                         "errmgr:base:update_state() %s) "
+                         "------- %s state updated for process %s",
+                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
+                         (NULL == name) ? "App. Process" : (name->jobid == ORTE_PROC_MY_HNP->jobid ? "Daemon" : "App. Process"),
+                         (NULL == name) ? "NULL" : ORTE_NAME_PRINT(name)));
     
     stack_state  = ORTE_ERRMGR_STACK_STATE_NONE;
     stack_state |= ORTE_ERRMGR_STACK_STATE_JOB_ABORT;
@@ -101,7 +95,7 @@ int orte_errmgr_base_update_state(orte_jobid_t job,
         }
     }
     
-    return ORTE_SUCCESS;
+    return rc;
 }
 
 int orte_errmgr_base_abort(int error_code, char *fmt, ...)

orte/mca/errmgr/base/errmgr_base_open.c

@@ -79,12 +79,6 @@ int orte_errmgr_base_open(void)
     
     orte_errmgr_base.output = opal_output_open(NULL);
 
-    /*
-     * A flag to indicate that orterun is shutting down, so skip the recovery
-     * logic.
-     */
-    orte_errmgr_base.shutting_down = false;
-
     /*
      * Open up all available components
      */

orte/mca/errmgr/base/errmgr_private.h

@@ -42,7 +42,6 @@ BEGIN_C_DECLS
 /* define a struct to hold framework-global values */
 typedef struct {
     int output;
-    bool shutting_down;
     opal_pointer_array_t modules;
     bool initialized;
 } orte_errmgr_base_t;

orte/mca/errmgr/hnp/errmgr_hnp.c

@@ -132,7 +132,7 @@ static int update_state(orte_jobid_t job,
     /*
      * if orte is trying to shutdown, just let it
      */
-    if (orte_errmgr_base.shutting_down) {
+    if (orte_finalizing) {
         return ORTE_SUCCESS;
     }
 
@@ -235,6 +235,18 @@ static int update_state(orte_jobid_t job,
                     hnp_abort(jdata->jobid, exit_code);
                 }
                 break;
+            case ORTE_JOB_STATE_COMM_FAILED:
+                /* order all local procs for this job to be killed */
+                killprocs(jdata->jobid, ORTE_VPID_WILDCARD);
+                check_job_complete(jdata);  /* set the local proc states */
+                /* the job object for this job will have been NULL'd
+                 * in the array if the job was solely local. If it isn't
+                 * NULL, then we need to tell everyone else to die
+                 */
+                if (NULL != (jdata = orte_get_job_data_object(job))) {
+                    hnp_abort(jdata->jobid, exit_code);
+                }
+                break;
 
             default:
                 break;
@@ -253,7 +265,6 @@ static int update_state(orte_jobid_t job,
         case ORTE_PROC_STATE_ABORTED:
         case ORTE_PROC_STATE_ABORTED_BY_SIG:
         case ORTE_PROC_STATE_TERM_WO_SYNC:
-        case ORTE_PROC_STATE_COMM_FAILED:
             if (jdata->enable_recovery) {
                 /* is this a local proc */
                 if (NULL != (child = proc_is_local(proc))) {
@@ -327,6 +338,12 @@ static int update_state(orte_jobid_t job,
             }
             break;
             
+        case ORTE_PROC_STATE_COMM_FAILED:
+            /* is this to a daemon? */
+            /* relocate its processes */
+            /* attempt to restart? */
+            break;
+
         default:
             break;
     }

orte/mca/errmgr/orted/errmgr_orted.c

@@ -78,7 +78,7 @@ static int ft_event(int state);
 
 
 /******************
- * ORCM module
+ * ORTED module
  ******************/
 orte_errmgr_base_module_t orte_errmgr_orted_module = {
     init,
@@ -123,7 +123,13 @@ static int update_state(orte_jobid_t job,
     /*
      * if orte is trying to shutdown, just let it
      */
-    if (orte_errmgr_base.shutting_down) {
+    if (orte_finalizing) {
+        return ORTE_SUCCESS;
+    }
+    
+    /* if this is a heartbeat failure, let the HNP handle it */
+    if (ORTE_JOB_STATE_HEARTBEAT_FAILED == jobstate ||
+        ORTE_PROC_STATE_HEARTBEAT_FAILED == state) {
         return ORTE_SUCCESS;
     }
     

orte/mca/ess/base/ess_base_std_app.c

@@ -44,6 +44,7 @@
 #include "orte/mca/odls/odls_types.h"
 #include "orte/mca/plm/plm.h"
 #include "orte/mca/filem/base/base.h"
+#include "orte/mca/errmgr/base/base.h"
 #if OPAL_ENABLE_FT_CR == 1
 #include "orte/mca/snapc/base/base.h"
 #endif
@@ -65,6 +66,18 @@ int orte_ess_base_app_setup(void)
     int ret;
     char *error = NULL;
 
+    /* setup the errmgr */
+    if (ORTE_SUCCESS != (ret = orte_errmgr_base_open())) {
+        ORTE_ERROR_LOG(ret);
+        error = "orte_errmgr_base_open";
+        goto error;
+    }
+    if (ORTE_SUCCESS != (ret = orte_errmgr_base_select())) {
+        ORTE_ERROR_LOG(ret);
+        error = "orte_errmgr_base_select";
+        goto error;
+    }
+    
     /* Setup the communication infrastructure */
     
     /* Runtime Messaging Layer */

orte/mca/oob/tcp/oob_tcp_peer.c

@@ -60,7 +60,6 @@
 #include "orte/util/name_fns.h"
 #include "orte/runtime/orte_globals.h"
 #include "orte/mca/errmgr/errmgr.h"
-#include "orte/mca/routed/routed.h"
 #include "orte/mca/ess/ess.h"
 #include "orte/mca/notifier/notifier.h"
 #include "orte/runtime/orte_wait.h"
@@ -597,12 +596,12 @@ void mca_oob_tcp_peer_close(mca_oob_tcp_peer_t* peer)
             peer->peer_state);
     }
 
-    mca_oob_tcp_peer_shutdown(peer);
-
-    /* inform the routed framework that we have lost a connection so
+    /* inform the ERRMGR framework that we have lost a connection so
      * it can decide if this is important, what to do about it, etc.
      */
-    if (ORTE_SUCCESS != orte_routed.route_lost(&peer->peer_name)) {
+    if (ORTE_ERR_UNRECOVERABLE == orte_errmgr.update_state(peer->peer_name.jobid, ORTE_JOB_STATE_COMM_FAILED,
+                                                           &peer->peer_name, ORTE_PROC_STATE_COMM_FAILED,
+                                                           ORTE_ERROR_DEFAULT_EXIT_CODE)) {
         /* Should free the peer lock before we abort so we don't 
          * get stuck in the orte_wait_kill when receiving messages in the 
          * tcp OOB

orte/tools/orterun/orterun.c

@@ -1228,8 +1228,7 @@ static void abort_exit_callback(int fd, short ign, void *arg)
          * procedure.
          */
         orte_enable_recovery             = false;
-        orte_errmgr_base.shutting_down   = true;
-
+        
         /* terminate the orteds - they will automatically kill
          * their local procs
          */

orte/util/error_strings.c

@@ -129,6 +129,10 @@ const char *orte_err2str(int errnum)
     case ORTE_ERR_RELOCATE_LIMIT_EXCEEDED:
         retval = "Limit on number of process relocations was exceeded";
         break;
+    case ORTE_ERR_UNRECOVERABLE:
+        retval = "Unrecoverable error";
+        break;
+
             
     default:
         retval = NULL;