Initial support for new hostbin startup mode

hpc · Apr 1, 2014 · 1dd436a · 1dd436a
1 parent 4c83074
commit 1dd436a
Show file tree

Hide file tree

Showing 33 changed files with 1,512 additions and 125 deletions.
diff --git a/config.h.in b/config.h.in
@@ -51,6 +51,9 @@
 /* Define to 1 if you have the <unistd.h> header file. */
 #undef HAVE_UNISTD_H
 
+/* Hostbin executable */
+#undef HOSTBIN_PATH
+
 /* Number of bytes in keys for gcrypt */
 #undef KEY_SIZE_BYTES
 

diff --git a/configure b/configure
@@ -969,6 +969,7 @@ with_testrm
 with_default_port
 with_localstorage
 enable_bluegene
+with_hostbin
 enable_pipes
 enable_socket
 enable_shmem
@@ -1679,6 +1680,7 @@ Optional Packages:
                           'slurm' and 'flux')
   --with-default-port=NUM TCP Port for Spindle server communication
   --with-localstorage=DIR Directory on back-ends for storing relocated files
+  --with-hostbin=EXE      Executable for returning host lists for jobs
   --with-python-prefix=STR
                           List of directories containing python installs
   --with-usage-logging=FILE
@@ -5131,13 +5133,13 @@ if test "${lt_cv_nm_interface+set}" = set; then
 else
   lt_cv_nm_interface="BSD nm"
   echo "int some_variable = 0;" > conftest.$ac_ext
-  (eval echo "\"\$as_me:5134: $ac_compile\"" >&5)
+  (eval echo "\"\$as_me:5136: $ac_compile\"" >&5)
   (eval "$ac_compile" 2>conftest.err)
   cat conftest.err >&5
-  (eval echo "\"\$as_me:5137: $NM \\\"conftest.$ac_objext\\\"\"" >&5)
+  (eval echo "\"\$as_me:5139: $NM \\\"conftest.$ac_objext\\\"\"" >&5)
   (eval "$NM \"conftest.$ac_objext\"" 2>conftest.err > conftest.out)
   cat conftest.err >&5
-  (eval echo "\"\$as_me:5140: output\"" >&5)
+  (eval echo "\"\$as_me:5142: output\"" >&5)
   cat conftest.out >&5
   if $GREP 'External.*some_variable' conftest.out > /dev/null; then
     lt_cv_nm_interface="MS dumpbin"
@@ -6343,7 +6345,7 @@ ia64-*-hpux*)
   ;;
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 6346 "configure"' > conftest.$ac_ext
+  echo '#line 6348 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -8934,11 +8936,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:8937: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:8939: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:8941: \$? = $ac_status" >&5
+   echo "$as_me:8943: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -9273,11 +9275,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:9276: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:9278: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:9280: \$? = $ac_status" >&5
+   echo "$as_me:9282: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -9378,11 +9380,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:9381: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:9383: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:9385: \$? = $ac_status" >&5
+   echo "$as_me:9387: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -9433,11 +9435,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:9436: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:9438: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:9440: \$? = $ac_status" >&5
+   echo "$as_me:9442: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -12236,7 +12238,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12239 "configure"
+#line 12241 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12332,7 +12334,7 @@ else
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<_LT_EOF
-#line 12335 "configure"
+#line 12337 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -14352,11 +14354,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:14355: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14357: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:14359: \$? = $ac_status" >&5
+   echo "$as_me:14361: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -14451,11 +14453,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:14454: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14456: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:14458: \$? = $ac_status" >&5
+   echo "$as_me:14460: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -14503,11 +14505,11 @@ else
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:14506: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14508: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:14510: \$? = $ac_status" >&5
+   echo "$as_me:14512: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -16076,6 +16078,17 @@ if test "x$OS_BUILD" == "x"; then
   OS_BUILD=linux
 fi
 
+#Startup type
+
+# Check whether --with-hostbin was given.
+if test "${with_hostbin+set}" = set; then
+  withval=$with_hostbin;
+cat >>confdefs.h <<_ACEOF
+#define HOSTBIN_PATH "${withval}"
+_ACEOF
+
+fi
+
 
 #Runmode detection (pipe/socket or cobo/msocket communications)
 if test "x$OS_BUILD" == "xlinux"; then

diff --git a/configure.common.ac b/configure.common.ac
@@ -21,6 +21,10 @@ if test "x$OS_BUILD" == "x"; then
   OS_BUILD=linux
 fi
 
+#Startup type
+AC_ARG_WITH(hostbin,
+            [AS_HELP_STRING([--with-hostbin=EXE],[Executable for returning host lists for jobs])],
+            [AC_DEFINE_UNQUOTED([HOSTBIN_PATH],"[${withval}]",[Hostbin executable])],)
 
 #Runmode detection (pipe/socket or cobo/msocket communications)
 if test "x$OS_BUILD" == "xlinux"; then

diff --git a/src/client/beboot/spindle_bootstrap.c b/src/client/beboot/spindle_bootstrap.c
@@ -21,6 +21,8 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 #include <string.h>
 #include <assert.h>
 #include <signal.h>
+#include <fcntl.h>
+#include <sys/wait.h>
 #include <sys/types.h>
 #include <sys/stat.h>
 
@@ -36,7 +38,7 @@ Place, Suite 330, Boston, MA 02111-1307 USA
 #if !defined(LIBEXECDIR)
 #error Expected to be built with libdir defined
 #endif
-
+char spindle_daemon[] = LIBEXECDIR "/spindle_be";
 
 int ldcsid;
 static int rankinfo[4]={-1,-1,-1,-1};
@@ -46,6 +48,7 @@ static char **cmdline;
 static char *executable;
 static char *client_lib;
 static char *opts_s;
+static char **daemon_args;
 
 unsigned long opts;
 
@@ -98,19 +101,71 @@ static void setup_environment()
 
 static int parse_cmdline(int argc, char *argv[])
 {
+   int i, daemon_arg_count;
    if (argc < 4)
       return -1;
+
+   i = 1;
+   if (strcmp(argv[1], "-daemon_args") == 0) {
+      debug_printf("Parsing daemon args out of bootstrap launch line\n");
+      daemon_arg_count = atoi(argv[2]);
+      daemon_args = malloc(sizeof(char *) * (daemon_arg_count + 2));
+      daemon_args[0] = spindle_daemon;
+      for (i = 4; i < 3 + daemon_arg_count; i++)
+         daemon_args[i - 3] = argv[i];
+      daemon_args[i - 3] = NULL;
+   }
 
-   location = argv[1];
-   number_s = argv[2];
+   location = argv[i++];
+   number_s = argv[i++];
    number = atoi(number_s);
-   opts_s = argv[3];
+   opts_s = argv[i++];
    opts = atoi(opts_s);
-   cmdline = argv + 4;
+   cmdline = argv + i;
+   assert(i < argc);
 
    return 0;
 }
 
+static void launch_daemon(char *location)
+{
+   /*grand-child fork, then execv daemon.  By grand-child forking we ensure that
+     the app won't get confused by seeing an unknown process as a child. */
+   pid_t child, gchild;
+   int status;
+   int fd;
+   char unique_file[MAX_PATH_LEN+1];
+   char buffer[32];
+
+   snprintf(unique_file, MAX_PATH_LEN, "%s/spindle_daemon_pid", location);
+   unique_file[MAX_PATH_LEN] = '\0';
+   fd = open(unique_file, O_CREAT | O_EXCL | O_WRONLY, 0600);
+   if (fd == -1) {
+      debug_printf("Not starting daemon -- %s already exists\n", unique_file);
+      return;
+   }
+   debug_printf("Client is spawning daemon\n");
+
+   child = fork();
+   if (child == 0) {
+      gchild = fork();
+      if (gchild != 0) {
+         snprintf(buffer, sizeof(buffer), "%d\n", getpid());
+         write(fd, buffer, strlen(buffer));
+         close(fd);
+         exit(0);
+      }
+      close(fd);
+      execv(spindle_daemon, daemon_args);
+      fprintf(stderr, "Spindle error: Could not execv daemon %s\n", daemon_args[0]);
+      exit(-1);
+   }
+   else if (child > 0) {
+      close(fd);
+      waitpid(child, &status, 0);
+   }
+}
+
 static void get_executable()
 {
    if (!(opts & OPT_RELOCAOUT)) {
@@ -129,8 +184,6 @@ static void get_executable()
       debug_printf("Relocated executable %s to %s\n", *cmdline, executable);
       chmod(executable, 0700);
    }
-
-
 }
 
 static void adjust_script()
@@ -226,6 +279,10 @@ int main(int argc, char *argv[])
       return -1;
    }
    location = realize(location);
+
+   if (daemon_args) {
+      launch_daemon(location);
+   }
 
    result = establish_connection();
    if (result == -1) {

diff --git a/src/client/config.h.in b/src/client/config.h.in
@@ -45,6 +45,9 @@
 /* Define to 1 if you have the <unistd.h> header file. */
 #undef HAVE_UNISTD_H
 
+/* Hostbin executable */
+#undef HOSTBIN_PATH
+
 /* Number of bytes in keys for gcrypt */
 #undef KEY_SIZE_BYTES