Add slurm_epilog and slurm_prolog for bluegene systems to distribution

(moved from local mpirun distribution).

NEWS

@@ -11,6 +11,8 @@ documents those changes that are of interest to users and admins.
  -- Fix squeue output format for "%s" (node select data).
  -- Queue job requesting more resources than exist in a partition if 
     that partition's state is DOWN (rather than just abort it).
+ -- Add prolog/epilog for bluegene to code base (moved from mpirun in CVS)
+ -- Add prolog, epilog and bluegene.conf.example to bluegene RMP
 
 * Changes in SLURM 0.4.4
 ========================

slurm.spec.in

@@ -215,6 +215,8 @@ rm -rf $RPM_BUILD_ROOT
 %files bluegene
 %defattr(-,root,root)
 %doc etc/bluegene.conf.example
+%{_bindir}/epilog
+%{_bindir}/prolog
 %{_libdir}/slurm/select_bluegene.so
 %config(noreplace) /etc/slurm/bluegene.conf
 #############################################################################
@@ -267,6 +269,7 @@ fi
 %changelog
 * Thu Feb 24 2005 Morris Jette <jette1@llnl.gov>
 - added bluegene.conf.example to distribution
+- added epilog and prolog to bluegene package
 * Thu Feb 10 2005 Morris Jette <jette1@llnl.gov>
 - disable rpm build termination if unpackaged files found (suse rpm bug)
 * Wed Jan 02 2005 Morris Jette <jette1@llnl.gov>

src/plugins/select/bluegene/Makefile.am

@@ -23,3 +23,7 @@ select_bluegene_la_LDFLAGS = $(SO_LDFLAGS) $(PLUGIN_FLAGS)
 
 select_bluegene_la_LIBADD  = $(top_builddir)/src/partition_allocator/libpartition_allocator.la 
 
+bin_PROGRAMS = slurm_prolog slurm_epilog
+
+slurm_prolog_SOURCES = slurm_prolog.c
+slurm_epilog_SOURCES = slurm_epilog.c

src/plugins/select/bluegene/slurm_epilog.c

+/*****************************************************************************\
+ *  slurm_epilog.c - Wait until the specified partition is no longer owned by
+ *	this user. This is executed via SLURM to synchronize the user's job 
+ *	execution with slurmctld configuration of partitions.
+ *
+ * NOTE: execute "/bgl/BlueLight/ppcfloor/bglsys/bin/db2profile" first
+ *****************************************************************************
+ *  Copyright (C) 2004 The Regents of the University of California.
+ *  Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
+ *  Written by Morris Jette <jette1@llnl.gov>
+ *  UCRL-CODE-2002-040.
+ *
+ *  This file is part of SLURM, a resource management program.
+ *  For details, see <http://www.llnl.gov/linux/slurm/>.
+ *
+ *  SLURM is free software; you can redistribute it and/or modify it under
+ *  the terms of the GNU General Public License as published by the Free
+ *  Software Foundation; either version 2 of the License, or (at your option)
+ *  any later version.
+ *
+ *  SLURM is distributed in the hope that it will be useful, but WITHOUT ANY
+ *  WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ *  FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ *  details.
+ *
+ *  You should have received a copy of the GNU General Public License along
+ *  with SLURM; if not, write to the Free Software Foundation, Inc.,
+ *  59 Temple Place, Suite 330, Boston, MA  02111-1307  USA.
+\*****************************************************************************/
+
+#if HAVE_CONFIG_H
+#  include "config.h"
+#endif
+#include <stdlib.h>
+
+#ifndef HAVE_BGL_FILES
+
+/* Just a stub, no synchronization to perform */
+int main(int argc, char *argv[])
+{
+        exit(0);
+}
+
+#else
+
+#include <errno.h>
+#include <pwd.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include "rm_api.h"
+
+#define _DEBUG 0
+#define MAX_RETRIES 20			/* max retry count in polling */
+#define POLL_SLEEP 3			/* retry interval in seconds  */
+#define MAX_DELAY (MAX_RETRIES * POLL_SLEEP)	/* time in seconds    */
+
+static void  _wait_part_owner(char *part_name, char *user_id);
+
+int main(int argc, char *argv[])
+{
+	char *job_id = NULL, *part_name = NULL, *user_id = NULL;
+
+	job_id = getenv("SLURM_JOBID");		/* get SLURM job ID */
+	if (!job_id)
+		fprintf(stderr, "SLURM_JOBID not set\n");
+
+	part_name = getenv("BGL_PARTITION_ID");	/* get partition ID */
+	if (!part_name) {
+		fprintf(stderr, "BGL_PARTITION_ID not set for job %s\n",
+			job_id);
+		exit(0);
+	}
+
+	user_id = getenv("SLURM_UID");          /* get SLURM user ID */
+	if (!user_id)
+		fprintf(stderr, "SLURM_UID not set for job %s\n", job_id);
+	else
+		_wait_part_owner(part_name, user_id);
+	exit(0);
+}
+
+static void  _wait_part_owner(char *part_name, char *user_id)
+{
+	uid_t target_uid;
+	int i, rc1, rc2;
+	rm_partition_t *part_ptr;
+	char *name;
+	struct passwd *pw_ent;
+
+	target_uid = atoi(user_id);
+
+#if _DEBUG
+	printf("Waiting for partition %s owner to change from %d.\n", 
+		part_name, target_uid);
+#endif
+
+	for (i=0; i<MAX_RETRIES; i++) {
+		if (i) {
+			sleep(POLL_SLEEP);
+#if _DEBUG
+			printf(".");
+#endif
+		}
+		if ((rc1 = rm_get_partition(part_name, &part_ptr)) != 
+				STATUS_OK) {
+			fprintf(stderr, "rm_get_partition(%s) errno=%d\n",
+				part_name, rc1);
+			return;
+		}
+		rc1 = rm_get_data(part_ptr, RM_PartitionUserName, &name);
+		rc2 = rm_free_partition(part_ptr);
+		if (rc1 != STATUS_OK) {
+			fprintf(stderr,
+				"rm_get_data(%s, RM_PartitionUserName) "
+				"errno=%d\n", part_name, rc1);
+			return;
+		}
+		if (rc2 != STATUS_OK)
+			fprintf(stderr, "rm_free_partition() errno=%d\n", rc2);
+
+		/* Now test this owner */
+		if (name[0] == '\0')
+			break;
+		if ((pw_ent = getpwnam(name)) == NULL) {
+			fprintf(stderr, "getpwnam(%s) errno=%d\n", part_name, 
+				errno);
+			continue;
+		}
+#if (_DEBUG > 1)
+		printf("\nowner = %s(%d)\n", name, pw_ent->pw_uid);
+#endif
+		if (pw_ent->pw_uid != target_uid)
+			break;
+	}
+
+#if _DEBUG
+	if (i >= MAX_RETRIES)
+		printf("\n");
+	else
+		printf("\nPartition %s owner is %d.\n", part_name, target_uid);
+#endif
+	if (i >= MAX_RETRIES)
+		fprintf(stderr, "Partition %s owner not changed (%s)\n", 
+			part_name, name);
+}
+
+#endif  /* HAVE_BGL_FILES */

src/plugins/select/bluegene/slurm_prolog.c

+/*****************************************************************************\
+ * slurm_ prolog.c - Wait until the specified partition is ready and owned by 
+ *	this user. This is executed via SLURM to synchronize the user's job 
+ *	execution with slurmctld configuration of partitions.
+ *
+ * NOTE: execute "/bgl/BlueLight/ppcfloor/bglsys/bin/db2profile" first
+ *****************************************************************************
+ *  Copyright (C) 2004 The Regents of the University of California.
+ *  Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
+ *  Written by Morris Jette <jette1@llnl.gov>
+ *  UCRL-CODE-2002-040.
+ *
+ *  This file is part of SLURM, a resource management program.
+ *  For details, see <http://www.llnl.gov/linux/slurm/>.
+ *
+ *  SLURM is free software; you can redistribute it and/or modify it under
+ *  the terms of the GNU General Public License as published by the Free
+ *  Software Foundation; either version 2 of the License, or (at your option)
+ *  any later version.
+ *
+ *  SLURM is distributed in the hope that it will be useful, but WITHOUT ANY
+ *  WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ *  FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ *  details.
+ *
+ *  You should have received a copy of the GNU General Public License along
+ *  with SLURM; if not, write to the Free Software Foundation, Inc.,
+ *  59 Temple Place, Suite 330, Boston, MA  02111-1307  USA.
+\*****************************************************************************/
+
+#if HAVE_CONFIG_H
+#  include "config.h"
+#endif
+#include <stdlib.h>
+
+#ifndef HAVE_BGL_FILES
+
+/* Just a stub, no synchronization to perform */
+int main(int argc, char *argv[])
+{
+	exit(0);
+}
+
+#else
+
+#include <errno.h>
+#include <pwd.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include "rm_api.h"
+
+#define _DEBUG 0
+
+/*
+ * Check the bglblock's status every POLL_SLEEP seconds. 
+ * Retry for a period of MIN_DELAY + (INCR_DELAY * base partition count).
+ * For example if MIN_DELAY=300 and INCR_DELAY=20, wait up to 428 seconds
+ * for a 16 base partition bglblock to ready (300 + 20 * 16).
+ */ 
+#define POLL_SLEEP 3			/* retry interval in seconds  */
+#define MIN_DELAY  300			/* time in seconds */
+#define INCR_DELAY 20			/* time in seconds per BP */
+int max_delay=MIN_DELAY;
+int cur_delay=0;
+
+static char *_part_state_str(rm_partition_state_t state);
+static void  _wait_part_ready(char *part_name);
+static void  _wait_part_owner(char *part_name, char *user_id);
+
+int main(int argc, char *argv[])
+{
+	char *job_id = NULL, *part_name = NULL, *user_id = NULL;
+
+	job_id = getenv("SLURM_JOBID");		/* get SLURM job ID */
+	if (!job_id)
+		fprintf(stderr, "SLURM_JOBID not set\n");
+
+	part_name = getenv("BGL_PARTITION_ID");	/* get partition ID */
+	if (!part_name) {
+		fprintf(stderr, "BGL_PARTITION_ID not set for job %s\n",
+			job_id);
+		exit(0);
+	}
+
+	_wait_part_ready(part_name);
+
+	user_id = getenv("SLURM_UID");          /* get SLURM user ID */
+	if (!user_id)
+		fprintf(stderr, "SLURM_UID not set for job %s\n", job_id);
+	else
+		_wait_part_owner(part_name, user_id);
+	exit(0);
+}
+
+static void _wait_part_ready(char *part_name)
+{
+	int i, rc1, rc2;
+	rm_partition_t *part_ptr;
+	rm_partition_state_t state;
+	int is_ready = 0;
+
+#if _DEBUG
+	printf("Waiting for partition %s to become ready.", part_name);
+#endif
+	for (i=0; (cur_delay < max_delay); i++) {
+		if (i) {
+			sleep(POLL_SLEEP);
+			cur_delay += POLL_SLEEP;
+#if _DEBUG
+			printf(".");
+#endif
+		}
+		rc1 = rm_get_partition(part_name, &part_ptr);
+		if (rc1 == PARTITION_NOT_FOUND)
+			continue;	/* wait for creation */
+		if (rc1 != STATUS_OK) {
+			fprintf(stderr, "rm_get_partition(%s) errno=%d\n", 
+				part_name, rc1);
+			return;
+		}
+
+		if (max_delay == MIN_DELAY) {
+			int bp;
+			rc1 = rm_get_data(part_ptr, RM_PartitionBPNum, &bp);
+			if (rc1 != STATUS_OK)
+				fprintf(stderr, "rm_get_data(%s, "
+					"RM_PartitionBPNum) errno=%d\n",
+					part_name, rc1);
+			else {
+				max_delay += (INCR_DELAY * bp);
+				if (max_delay == MIN_DELAY)
+					max_delay++;	/* avoid re-test */
+			}
+		}
+		rc1 = rm_get_data(part_ptr, RM_PartitionState, &state);
+		rc2 = rm_free_partition(part_ptr);
+		if (rc1 != STATUS_OK) {
+			fprintf(stderr, 
+				"rm_get_data(%s, RM_PartitionState) errno=%d\n", 
+				part_name, rc1);
+			return;
+		}
+		if (rc2 != STATUS_OK)
+			fprintf(stderr, "rm_free_partition() errno=%d\n", rc2);
+#if (_DEBUG > 1)
+		printf("\nstate=%s\n",_part_state_str(state));
+#endif
+		if ((state == RM_PARTITION_READY) 
+		||  (state == RM_PARTITION_ERROR)) {
+			is_ready = 1;
+			break;
+		}
+	}
+#if _DEBUG
+	if (is_ready == 0)
+		printf("\n");
+	else
+     	   printf("\nPartition %s is ready.\n", part_name);
+#endif
+	if (is_ready == 0)
+		fprintf(stderr, "Partition state not ready (%s)\n",
+			_part_state_str(state));
+
+}
+
+static char *_part_state_str(rm_partition_state_t state)
+{
+	static char tmp[16];
+
+	switch (state) {
+		case RM_PARTITION_BUSY: 
+			return "RM_PARTITION_BUSY";
+		case RM_PARTITION_CONFIGURING:
+			return "RM_PARTITION_CONFIGURING";
+		case RM_PARTITION_DEALLOCATING:
+			return "RM_PARTITION_DEALLOCATING";
+		case RM_PARTITION_ERROR:
+			return "RM_PARTITION_ERROR";
+		case RM_PARTITION_FREE:
+			return "RM_PARTITION_FREE";
+		case RM_PARTITION_NAV:
+			return "RM_PARTITION_NAV";
+		case RM_PARTITION_READY:
+			return "RM_PARTITION_READY";
+		default:
+			snprintf(tmp, sizeof(tmp), "%d", state);
+			return tmp;
+	}
+}
+
+/* Partition owner should be set when partition is ready, don't 
+ * have long delays */
+static void  _wait_part_owner(char *part_name, char *user_id)
+{
+	uid_t target_uid;
+	int i, rc1, rc2;
+	rm_partition_t *part_ptr;
+	char *name;
+	struct passwd *pw_ent;
+	int is_ready = 0;
+
+	target_uid = atoi(user_id);
+
+#if _DEBUG
+	printf("Waiting for partition %s owner to become %d.", part_name, 
+		target_uid);
+#endif
+
+	for (i=0; (cur_delay < max_delay); i++) {
+		if (i) {
+			sleep(POLL_SLEEP);
+			cur_delay += POLL_SLEEP;
+#if _DEBUG
+			printf(".");
+#endif
+		}
+		if ((rc1 = rm_get_partition(part_name, &part_ptr)) != 
+				STATUS_OK) {
+			fprintf(stderr, "rm_get_partition(%s) errno=%d\n",
+				part_name, rc1);
+			return;
+		}
+		rc1 = rm_get_data(part_ptr, RM_PartitionUserName, &name);
+		rc2 = rm_free_partition(part_ptr);
+		if (rc1 != STATUS_OK) {
+			fprintf(stderr,
+				"rm_get_data(%s, RM_PartitionUserName) "
+				"errno=%d\n", part_name, rc1);
+			return;
+		}
+		if (rc2 != STATUS_OK)
+			fprintf(stderr, "rm_free_partition() errno=%d\n", rc2);
+
+		/* Now test this owner */
+		if (name[0] == '\0')
+			continue;
+		if ((pw_ent = getpwnam(name)) == NULL) {
+			fprintf(stderr, "getpwnam(%s) errno=%d\n", part_name, 
+				errno);
+			continue;
+		}
+#if (_DEBUG > 1)
+		printf("\nowner = %s(%d)\n", name, pw_ent->pw_uid);
+#endif
+		if (pw_ent->pw_uid == target_uid) {
+			is_ready = 1;
+			break;
+		}
+	}
+
+#if _DEBUG
+	if (is_ready == 0)
+		printf("\n");
+	else
+		printf("\nPartition %s owner is %d.\n", part_name, target_uid);
+#endif
+	if (is_ready == 0)
+		fprintf(stderr, "Partition %s owner not changed (%s)\n", 
+			part_name, name);
+}
+
+#endif  /* HAVE_BGL_FILES */