diff --git a/NEWS b/NEWS
index d2a1d3fb1cd3288534b6d05576a05e22e9a282b1..dd820164c5193c37759dff2eb051e7f8b97d8f07 100644
--- a/NEWS
+++ b/NEWS
@@ -1,6 +1,10 @@
This file describes changes in recent versions of SLURM. It primarily
documents those changes that are of interest to users and admins.
+* Changes in SLURM 1.1.0-pre6
+=============================
+ -- Accounting works for aix systems use jobacct/aix
+
* Changes in SLURM 1.1.0-pre6
=============================
-- Added logic to "stat" a running job with sacct option -S use -j to specify
diff --git a/src/api/job_info.c b/src/api/job_info.c
index 7395dd66d17d9c2f276d1d1748bcc0008853bb83..4c65ceeb833374fea01d823704487acc5641a356 100644
--- a/src/api/job_info.c
+++ b/src/api/job_info.c
@@ -269,7 +269,7 @@ slurm_print_job_info ( FILE* out, job_info_t * job_ptr, int one_liner )
strncpy(time_str, "None", sizeof(time_str));
}
fprintf ( out, "SuspendTime=%s PreSusTime=%ld",
- time_str, job_ptr->pre_sus_time);
+ time_str, (long int)job_ptr->pre_sus_time);
/****** Line 13 (optional) ******/
select_g_sprint_jobinfo(job_ptr->select_jobinfo,
diff --git a/src/common/env.c b/src/common/env.c
index 97ca68e7ca3485a6406b56ba8862a98ae382edd6..c7cac5a28dc581ef43918cca0af28c74b7f4b518 100644
--- a/src/common/env.c
+++ b/src/common/env.c
@@ -502,7 +502,7 @@ int setup_env(env_t *env)
#ifdef HAVE_AIX
{
- char res_env[128], tmp_env[32];
+ char res_env[128];
char *debug_env = (char *)getenv("SLURM_LL_API_DEBUG");
int debug_num = 0;
diff --git a/src/plugins/checkpoint/aix/checkpoint_aix.c b/src/plugins/checkpoint/aix/checkpoint_aix.c
index 873ee38c7ce0af7f2fdf7e634d9fa743bade8b02..e0f18ae8c6c00570c48060304937105f71ef9b7a 100644
--- a/src/plugins/checkpoint/aix/checkpoint_aix.c
+++ b/src/plugins/checkpoint/aix/checkpoint_aix.c
@@ -412,7 +412,7 @@ static void *_ckpt_agent_thr(void *arg)
iter = list_iterator_create(ckpt_timeout_list);
slurm_mutex_lock(&ckpt_agent_mutex);
/* look for and process any timeouts */
- while (rec = list_next(iter)) {
+ while ((rec = list_next(iter))) {
if (rec->end_time > now)
continue;
info("checkpoint timeout for %u.%u",
@@ -479,7 +479,7 @@ static void _ckpt_dequeue_timeout(uint32_t job_id, uint32_t step_id,
if (!ckpt_timeout_list)
goto fini;
iter = list_iterator_create(ckpt_timeout_list);
- while (rec = list_next(iter)) {
+ while ((rec = list_next(iter))) {
if ((rec->job_id != job_id) || (rec->step_id != step_id)
|| (start_time && (rec->start_time != start_time)))
continue;
diff --git a/src/plugins/jobacct/aix/jobacct_aix.c b/src/plugins/jobacct/aix/jobacct_aix.c
index cdf23807b32c2c166b6d3111ad541d3199c9155b..c9394869d4faebd4a0dd0ddb8439c334da6b0e08 100644
--- a/src/plugins/jobacct/aix/jobacct_aix.c
+++ b/src/plugins/jobacct/aix/jobacct_aix.c
@@ -30,6 +30,12 @@
#include "src/plugins/jobacct/common/jobacct_common.h"
+#ifdef HAVE_AIX
+#include <procinfo.h>
+#include <sys/types.h>
+#define NPROCS 5000
+#endif
+
/*
* These variables are required by the generic plugin interface. If they
@@ -61,28 +67,37 @@
* matures.
*/
const char plugin_name[] =
-"Job accounting LINUX plugin";
-const char plugin_type[] = "jobacct/linux";
+"Job accounting AIX plugin";
+const char plugin_type[] = "jobacct/aix";
const uint32_t plugin_version = 100;
/* Other useful declarations */
-#if 0
+#ifdef HAVE_AIX
typedef struct prec { /* process record */
- pid_t pid;
- pid_t ppid;
- int psize; /* maxrss */
- int vsize; /* max virtual size */
+ pid_t pid;
+ pid_t ppid;
+ int usec; /* user cpu time */
+ int ssec; /* system cpu time */
+ int pages; /* pages */
+ float rss; /* maxrss */
+ float vsize; /* max virtual size */
} prec_t;
static int freq = 0;
-static List prec_list = NULL;
/* Finally, pre-define all the routines. */
-static void _get_offspring_data(prec_t *ancestor, pid_t pid);
-static void _get_process_data(pid_t pid);
-static int _get_process_data_line(FILE *in, prec_t *prec);
+static void _get_offspring_data(List prec_list, prec_t *ancestor, pid_t pid);
+static void _get_process_data();
static void *_watch_tasks(void *arg);
static void _destroy_prec(void *object);
+
+/* system call to get process table */
+extern int getprocs(struct procsinfo *procinfo, int, struct fdsinfo *,
+ int, pid_t *, int);
+ /* procinfo: pointer to array of procinfo struct */
+ /* nproc: number of user procinfo struct */
+ /* sizproc: size of expected procinfo structure */
+
#endif
/*
* The following routine is called by the slurmd mainline
@@ -181,29 +196,22 @@ int jobacct_p_startpoll(int frequency)
{
int rc = SLURM_SUCCESS;
-#if 0
pthread_attr_t attr;
pthread_t _watch_tasks_thread_id;
-#endif
- debug("jobacct AIX plugin loaded");
- return rc;
-#if 0
- /* FIXME!!!!!!!!!!!!!!!!!!!!
- This was written for linux systems doesn't to anything on AIX */
-
- /* Parse the JobAcctParameters */
+ debug("jobacct AIX plugin loaded");
debug("jobacct: frequency = %d", frequency);
fini = false;
if (frequency == 0) { /* don't want dynamic monitoring? */
- debug2("jobacct LINUX dynamic logging disabled");
+ debug2("jobacct AIX dynamic logging disabled");
return rc;
}
freq = frequency;
+ task_list = list_create(common_free_jobacct);
/* create polling thread */
slurm_attr_init(&attr);
@@ -217,9 +225,8 @@ int jobacct_p_startpoll(int frequency)
frequency = 0;
}
else
- debug3("jobacct LINUX dynamic logging enabled");
+ debug3("jobacct AIX dynamic logging enabled");
slurm_attr_destroy(&attr);
-#endif
return rc;
}
@@ -248,7 +255,8 @@ void jobacct_p_suspendpoll()
common_suspendpoll();
}
-#if 0
+#ifdef HAVE_AIX
+
/*
* _get_offspring_data() -- collect memory usage data for the offspring
*
@@ -256,7 +264,8 @@ void jobacct_p_suspendpoll()
* usage data to the ancestor's <prec> record. Recurse to gather data
* for *all* subsequent generations.
*
- * IN: ancestor The entry in precTable[] to which the data
+ * IN: prec_list list of prec's
+ * ancestor The entry in prec_list to which the data
* should be added. Even as we recurse, this will
* always be the prec for the base of the family
* tree.
@@ -269,9 +278,27 @@ void jobacct_p_suspendpoll()
*
* THREADSAFE! Only one thread ever gets here.
*/
-static void _get_offspring_data(prec_t *ancestor, pid_t pid)
+static void _get_offspring_data(List prec_list, prec_t *ancestor, pid_t pid)
{
-
+ ListIterator itr;
+ prec_t *prec = NULL;
+
+ itr = list_iterator_create(prec_list);
+ while((prec = list_next(itr))) {
+ if (prec->ppid == pid) {
+ _get_offspring_data(prec_list, ancestor, prec->pid);
+ debug2("adding %d to %d rss = %f vsize = %f",
+ prec->pid, ancestor->pid,
+ prec->rss, prec->vsize);
+ ancestor->usec += prec->usec;
+ ancestor->ssec += prec->ssec;
+ ancestor->pages += prec->pages;
+ ancestor->rss += prec->rss;
+ ancestor->vsize += prec->vsize;
+ }
+ }
+ list_iterator_destroy(itr);
+
return;
}
@@ -289,31 +316,89 @@ static void _get_offspring_data(prec_t *ancestor, pid_t pid)
* is a Linux-style stat entry. We disregard the data if they look
* wrong.
*/
-static void _get_process_data(pid_t pid)
+static void _get_process_data()
{
+ struct procsinfo proc;
+ int pid = 0;
+ static int processing = 0;
+ prec_t *prec = NULL;
+ struct jobacctinfo *jobacct = NULL;
+ List prec_list = NULL;
+ ListIterator itr;
+ ListIterator itr2;
+
+ if(processing) {
+ debug("already running, returning");
+ return;
+ }
+
+ processing = 1;
+ prec_list = list_create(_destroy_prec);
+ /* get the whole process table */
+ while(getprocs(&proc, sizeof(proc), 0, 0, &pid, 1) == 1) {
+ prec = xmalloc(sizeof(prec_t));
+ list_append(prec_list, prec);
+ prec->pid = proc.pi_pid;
+ prec->ppid = proc.pi_ppid;
+ prec->usec = proc.pi_ru.ru_utime.tv_sec +
+ proc.pi_ru.ru_utime.tv_usec * 1e-6;
+ prec->ssec = proc.pi_ru.ru_stime.tv_sec +
+ proc.pi_ru.ru_stime.tv_usec * 1e-6;
+ prec->pages = proc.pi_majflt;
+ prec->rss = (proc.pi_trss + proc.pi_drss) * 4;
+ prec->rss *= 1024;
+ prec->vsize = (proc.pi_tsize / 1024);
+ prec->vsize += (proc.pi_dvm * 4);
+ prec->vsize *= 1024;
+ /* debug("vsize = %f = %d/1024+%d", */
+/* prec->vsize, proc.pi_tsize, proc.pi_dvm * 4); */
+ }
+ if(!list_count(prec_list))
+ goto finished;
+
+ slurm_mutex_lock(&jobacct_lock);
+ if(!task_list || !list_count(task_list)) {
+ slurm_mutex_unlock(&jobacct_lock);
+ goto finished;
+ }
+ itr = list_iterator_create(task_list);
+ while((jobacct = list_next(itr))) {
+ itr2 = list_iterator_create(prec_list);
+ while((prec = list_next(itr2))) {
+ //debug2("pid %d ? %d", prec->ppid, jobacct->pid);
+ if (prec->pid == jobacct->pid) {
+ /* find all my descendents */
+ _get_offspring_data(prec_list, prec,
+ prec->pid);
+
+ /* tally their usage */
+ jobacct->max_rss = jobacct->tot_rss =
+ MAX(jobacct->max_rss, (int)prec->rss);
+ jobacct->max_vsize = jobacct->tot_vsize =
+ MAX(jobacct->max_vsize,
+ (int)prec->vsize);
+ jobacct->max_pages = jobacct->tot_pages
+ = MAX(jobacct->max_pages, prec->pages);
+ jobacct->min_cpu = jobacct->tot_cpu =
+ (prec->usec + prec->ssec);
+ debug2("%d size now %d %d time %d",
+ jobacct->pid, jobacct->max_rss,
+ jobacct->max_vsize, jobacct->tot_cpu);
+
+ break;
+ }
+ }
+ list_iterator_destroy(itr2);
+ }
+ list_iterator_destroy(itr);
+ slurm_mutex_unlock(&jobacct_lock);
+finished:
+ list_destroy(prec_list);
+ processing = 0;
return;
}
-/* _get_process_data_line() - get line of data from /proc/<pid>/stat
- *
- * IN: in - input file channel
- * OUT: prec - the destination for the data
- *
- * RETVAL: ==0 - no valid data
- * !=0 - data are valid
- *
- * Note: It seems a bit wasteful to do all those atoi() and
- * atol() conversions that are implicit in the scanf(),
- * but they help to ensure that we really are looking at the
- * expected type of record.
- */
-static int _get_process_data_line(FILE *in, prec_t *prec)
-{
- /* discardable data */
-
- return 1;
-}
/* _watch_tasks() -- monitor slurm jobs and track their memory usage
*
@@ -323,10 +408,9 @@ static int _get_process_data_line(FILE *in, prec_t *prec)
static void *_watch_tasks(void *arg)
{
- pid_t pid = getpid();
while(!fini) { /* Do this until slurm_jobacct_task_exit() stops us */
if(!suspended) {
- _get_process_data(pid); /* Update the data */
+ _get_process_data(); /* Update the data */
}
sleep(freq);
}
diff --git a/src/plugins/jobacct/common/common_slurmctld.c b/src/plugins/jobacct/common/common_slurmctld.c
index 0607c16dc2d9707e4f73012e0b473633f31a7535..d0f460128619561c426933532b4bc020a7b8f62e 100644
--- a/src/plugins/jobacct/common/common_slurmctld.c
+++ b/src/plugins/jobacct/common/common_slurmctld.c
@@ -371,13 +371,18 @@ extern int common_step_complete_slurmctld(struct step_record *step)
#endif
/* figure out the ave of the totals sent */
if(step->num_tasks > 0) {
- ave_vsize = jobacct->tot_vsize/step->num_tasks;
- ave_rss = jobacct->tot_rss/step->num_tasks;
- ave_pages = jobacct->tot_pages/step->num_tasks;
- ave_cpu = jobacct->tot_cpu/step->num_tasks;
+ ave_vsize = jobacct->tot_vsize;
+ ave_vsize /= step->num_tasks;
+ ave_rss = jobacct->tot_rss;
+ ave_rss /= step->num_tasks;
+ ave_pages = jobacct->tot_pages;
+ ave_pages /= step->num_tasks;
+ ave_cpu = jobacct->tot_cpu;
+ ave_cpu /= step->num_tasks;
ave_cpu /= 100;
}
- ave_cpu2 = jobacct->min_cpu/100;
+ ave_cpu2 = jobacct->min_cpu;
+ ave_cpu2 /= 100;
snprintf(buf, BUFFER_SIZE, _jobstep_format,
JOB_STEP,
diff --git a/src/plugins/jobacct/linux/jobacct_linux.c b/src/plugins/jobacct/linux/jobacct_linux.c
index dbe2b2803c9db0e09ea51ee69317dc627beac435..acff1f2b1d66415a3149056ac803ad273375105e 100644
--- a/src/plugins/jobacct/linux/jobacct_linux.c
+++ b/src/plugins/jobacct/linux/jobacct_linux.c
@@ -397,7 +397,7 @@ static void _get_process_data() {
while((jobacct = list_next(itr))) {
itr2 = list_iterator_create(prec_list);
while((prec = list_next(itr2))) {
- if (prec->ppid == jobacct->pid) {
+ if (prec->pid == jobacct->pid) {
/* find all my descendents */
_get_offspring_data(prec_list,
prec, prec->pid);
diff --git a/src/plugins/select/bluegene/block_allocator/block_allocator.c b/src/plugins/select/bluegene/block_allocator/block_allocator.c
index ac166789b2ee145ba7543902b33d8180f6bf5c99..f8bbe2c8569dccc9342151a04af733d5d05c99da 100644
--- a/src/plugins/select/bluegene/block_allocator/block_allocator.c
+++ b/src/plugins/select/bluegene/block_allocator/block_allocator.c
@@ -2738,13 +2738,13 @@ static int _find_x_path(List results, ba_node_t *ba_node,
ba_switch_t *curr_switch = NULL;
ba_switch_t *next_switch = NULL;
- int port_tar;
+ int port_tar = 0;
int source_port=0;
int target_port=0;
int broke = 0, not_first = 0;
int ports_to_try[2] = {3,5};
int *node_tar = NULL;
- int i;
+ int i = 0;
ba_node_t *next_node = NULL;
ba_node_t *check_node = NULL;
int highest_phys_x = geometry[X] - start[X];
@@ -3054,7 +3054,7 @@ static int _find_x_path2(List results, ba_node_t *ba_node,
int broke = 0, not_first = 0;
int ports_to_try[2] = {3,5};
int *node_tar = NULL;
- int i;
+ int i = 0;
ba_node_t *next_node = NULL;
ba_node_t *check_node = NULL;
diff --git a/src/plugins/switch/federation/federation.c b/src/plugins/switch/federation/federation.c
index 9a06b268c7b208ebcd3e941686b4534f0c42a823..8be66221d6e021d356730a3be4fac0238e2b624e 100644
--- a/src/plugins/switch/federation/federation.c
+++ b/src/plugins/switch/federation/federation.c
@@ -1454,10 +1454,10 @@ static int
_window_state_set(int adapter_cnt, fed_tableinfo_t *tableinfo,
char *hostname, int task_id, enum NTBL_RC state)
{
- fed_nodeinfo_t *node;
+ fed_nodeinfo_t *node = NULL;
fed_adapter_t *adapter = NULL;
- fed_window_t *window;
- NTBL *table;
+ fed_window_t *window = NULL;
+ NTBL *table = NULL;
int i, j;
bool adapter_found;
@@ -1668,7 +1668,7 @@ fed_build_jobinfo(fed_jobinfo_t *jp, hostlist_t hl, int nprocs,
{
int nnodes;
hostlist_iterator_t hi;
- char *host;
+ char *host = NULL;
int proc_cnt = 0;
int i, j;
fed_nodeinfo_t *node;
diff --git a/src/plugins/switch/federation/federation.h b/src/plugins/switch/federation/federation.h
index a3848ad5fff7cc0f9d63ea55d96a34db92cf7db0..9e2a1108a146dacd430901ca1ef021b3018ec6c3 100644
--- a/src/plugins/switch/federation/federation.h
+++ b/src/plugins/switch/federation/federation.h
@@ -66,6 +66,9 @@ enum {
#define FED_MAXADAPTERS 2
#define FED_LIBSTATE_LEN (1024 * 1024 * 1)
+int fed_slurmctld_init(void);
+int fed_slurmd_init(void);
+int fed_slurmd_step_init(void);
int fed_alloc_nodeinfo(fed_nodeinfo_t **nh);
int fed_build_nodeinfo(fed_nodeinfo_t *np, char *hostname);
char *fed_print_nodeinfo(fed_nodeinfo_t *np, char *buf, size_t size);
diff --git a/src/plugins/switch/federation/switch_federation.c b/src/plugins/switch/federation/switch_federation.c
index a4de81c9acf277da7e70595bef820d047b8b9888..3468808d7bcd3f73abf03608ba9829e1a341c433 100644
--- a/src/plugins/switch/federation/switch_federation.c
+++ b/src/plugins/switch/federation/switch_federation.c
@@ -35,6 +35,7 @@
#include <signal.h>
#include <sys/types.h>
#include <regex.h>
+#include <stdlib.h>
#include <slurm/slurm_errno.h>
#include "src/common/slurm_xlator.h"
@@ -222,7 +223,7 @@ int switch_p_libstate_restore ( char * dir_name, bool recover )
data = xmalloc(data_allocated);
while (1) {
data_read = read (state_fd, &data[data_size],
- FED_BUF_SIZE);
+ FED_BUF_SIZE);
if ((data_read < 0) && (errno == EINTR))
continue;
if (data_read < 0) {
diff --git a/src/sacct/options.c b/src/sacct/options.c
index 5efe7b0462c8488523ef27fdc17f015e92aa3d20..84013f1efd4649be7fcd9d17aaa853ae14d4a8a9 100644
--- a/src/sacct/options.c
+++ b/src/sacct/options.c
@@ -28,6 +28,7 @@
#include "src/common/read_config.h"
#include "sacct.h"
+#include <time.h>
typedef struct expired_rec { /* table of expired jobs */
uint32_t job;
@@ -122,13 +123,18 @@ int _cmp_jrec(const void *a1, const void *a2) {
*/
void _dump_header(acct_header_t header)
{
- printf("%u %s %d %d %d %d %s %s ",
+ struct tm ts;
+ gmtime_r(&header.timestamp, &ts);
+ printf("%u %s %04d%02d%02d%02d%02d%02d %d %s %s ",
header.jobnum,
header.partition,
+ 1900+(ts.tm_year),
+ 1+(ts.tm_mon),
+ ts.tm_mday,
+ ts.tm_hour,
+ ts.tm_min,
+ ts.tm_sec,
(int)header.job_start,
- (int)header.timestamp,
- header.uid,
- header.gid,
header.blockid, /* block id */
"-"); /* reserved 1 */
}
@@ -364,7 +370,7 @@ int decode_status_char(char *status)
char *decode_status_int(int status)
{
- switch(status) {
+ switch(status & ~JOB_COMPLETING) {
case JOB_PENDING:
return "PENDING"; /* we should never see this */
case JOB_RUNNING:
@@ -386,23 +392,33 @@ char *decode_status_int(int status)
default:
return "UNKNOWN";
}
- /* if (!strcasecmp(cs, "ca")) */
-/* return "CANCELLED"; */
-/* else if (strcasecmp(cs, "cd")==0) */
-/* return "COMPLETED"; */
-/* else if (strcasecmp(cs, "cg")==0) */
-/* return "COMPLETING"; /\* we should never see this *\/ */
-/* else if (strcasecmp(cs, "f")==0) */
-/* return "FAILED"; */
-/* else if (strcasecmp(cs, "nf")==0) */
-/* return "NODEFAILED"; */
-/* else if (strcasecmp(cs, "p")==0) */
-/* return "PENDING"; /\* we should never see this *\/ */
-/* else if (strcasecmp(cs, "r")==0) */
-/* return "RUNNING"; */
-/* else if (strcasecmp(cs, "to")==0) */
-/* return "TIMEDOUT"; */
-}
+}
+
+char *decode_status_int_abbrev(int status)
+{
+ switch(status & ~JOB_COMPLETING) {
+ case JOB_PENDING:
+ return "PD"; /* we should never see this */
+ case JOB_RUNNING:
+ return "R";
+ case JOB_SUSPENDED:
+ return "S";
+ case JOB_COMPLETE:
+ return "CD";
+ case JOB_CANCELLED:
+ return "CA";
+ case JOB_FAILED:
+ return "F";
+ case JOB_TIMEOUT:
+ return "TO";
+ case JOB_NODE_FAIL:
+ return "NF";
+ case JOB_END:
+ return "JOB_END";
+ default:
+ return "UNKNOWN";
+ }
+}
int get_data(void)
{
@@ -986,14 +1002,29 @@ endopt:
return;
}
+/* Note: do_dump() strives to present data in an upward-compatible
+ * manner so that apps written to use data from `sacct -d` in slurm
+ * v1.0 will continue to work in v1.1 and later.
+ *
+ * To help ensure this compatibility,
+ * a. The meaning of an existing field never changes
+ * b. New fields are appended to the end of a record
+ *
+ * The "numfields" field of the record can be used as a sub-version
+ * number, as it will never decrease for the life of the current
+ * record version number (currently 1). For example, if your app needs
+ * to use field 28, a record with numfields<28 is too old a version
+ * for you, while numfields>=28 will provide what you are expecting.
+ */
void do_dump(void)
{
ListIterator itr = NULL;
ListIterator itr_step = NULL;
job_rec_t *job = NULL;
step_rec_t *step = NULL;
- float tempf;
-
+ struct tm ts;
+ time_t finished;
+
itr = list_iterator_create(jobs);
while((job = list_next(itr))) {
if (!params.opt_dup)
@@ -1009,18 +1040,14 @@ void do_dump(void)
if (params.opt_uid>=0)
if (job->header.uid != params.opt_uid)
continue;
- if(job->sacct.min_cpu == NO_VAL)
+ if(job->sacct.min_cpu == (float)NO_VAL)
job->sacct.min_cpu = 0;
-
+
if(list_count(job->steps)) {
- tempf = job->sacct.ave_cpu/list_count(job->steps);
- job->sacct.ave_cpu = (uint32_t)tempf;
- tempf = job->sacct.ave_rss/list_count(job->steps);
- job->sacct.ave_rss = (uint32_t)tempf;
- tempf = job->sacct.ave_vsize/list_count(job->steps);
- job->sacct.ave_vsize = (uint32_t)tempf;
- tempf = job->sacct.ave_pages/list_count(job->steps);
- job->sacct.ave_pages = (uint32_t)tempf;
+ job->sacct.ave_cpu /= list_count(job->steps);
+ job->sacct.ave_rss /= list_count(job->steps);
+ job->sacct.ave_vsize /= list_count(job->steps);
+ job->sacct.ave_pages /= list_count(job->steps);
}
/* JOB_START */
@@ -1034,7 +1061,9 @@ void do_dump(void)
job->header.jobnum);
}
_dump_header(job->header);
- printf("JOB_START %s %d %d %d %s\n",
+ printf("JOB_START 1 16 %d %d %s %d %d %d %s\n",
+ job->header.uid,
+ job->header.gid,
job->jobname,
job->track_steps,
job->priority,
@@ -1050,12 +1079,14 @@ void do_dump(void)
step->exitcode=1;
}
_dump_header(step->header);
- printf("JOB_STEP %u %s %s ",
+ finished=step->header.job_start+step->elapsed;
+ gmtime_r(&finished, &ts);
+ printf("JOB_STEP 1 50 %u %04d%02d%02d%02d%02d%02d ",
step->stepnum,
- step->stepname,
- step->nodes);
+ 1900+(ts.tm_year), 1+(ts.tm_mon), ts.tm_mday,
+ ts.tm_hour, ts.tm_min, ts.tm_sec);
printf("%s %d %d %d %d ",
- decode_status_int(step->status),
+ decode_status_int_abbrev(step->status),
step->exitcode,
step->ntasks,
step->ncpus,
@@ -1068,7 +1099,7 @@ void do_dump(void)
(int)step->rusage.ru_stime.tv_sec,
(int)step->rusage.ru_stime.tv_usec);
printf("%d %d %d %d %d %d %d %d %d "
- "%d %d %d %d %d ",
+ "%d %d %d %d %d %d %d ",
(int)step->rusage.ru_maxrss,
(int)step->rusage.ru_ixrss,
(int)step->rusage.ru_idrss,
@@ -1082,13 +1113,14 @@ void do_dump(void)
(int)step->rusage.ru_msgrcv,
(int)step->rusage.ru_nsignals,
(int)step->rusage.ru_nvcsw,
- (int)step->rusage.ru_nivcsw);
- printf("%d %d %.2f %d %d %.2f "
- "%d %d %.2f %.2f %d %.2f\n",
+ (int)step->rusage.ru_nivcsw,
step->sacct.max_vsize,
+ step->sacct.max_rss);
+ /* Data added in Slurm v1.1 */
+ printf("%d %.2f %d %.2f %d %d %.2f "
+ "%.2f %d %.2f %s %s\n",
step->sacct.max_vsize_task,
step->sacct.ave_vsize,
- step->sacct.max_rss,
step->sacct.max_rss_task,
step->sacct.ave_rss,
step->sacct.max_pages,
@@ -1096,16 +1128,23 @@ void do_dump(void)
step->sacct.ave_pages,
step->sacct.min_cpu,
step->sacct.min_cpu_task,
- step->sacct.ave_cpu);
+ step->sacct.ave_cpu,
+ step->stepname,
+ step->nodes);
}
list_iterator_destroy(itr_step);
/* JOB_TERMINATED */
if (job->show_full) {
_dump_header(job->header);
- printf("JOB_TERMINATED %d ",
+ finished=job->header.job_start+job->elapsed;
+ gmtime_r(&finished, &ts);
+ printf("JOB_TERMINATED 1 50 %d ",
job->elapsed);
+ printf("%04d%02d%02d%02d%02d%02d ",
+ 1900+(ts.tm_year), 1+(ts.tm_mon), ts.tm_mday,
+ ts.tm_hour, ts.tm_min, ts.tm_sec);
printf("%s %d %d %d %d ",
- decode_status_int(job->status),
+ decode_status_int_abbrev(job->status),
job->exitcode,
job->ntasks,
job->ncpus,
@@ -1117,14 +1156,14 @@ void do_dump(void)
(int)job->rusage.ru_utime.tv_usec,
(int)job->rusage.ru_stime.tv_sec,
(int)job->rusage.ru_stime.tv_usec);
- printf("%d %d %d %d %d %d ",
+ printf("%d %d %d %d %d %d %d %d %d "
+ "%d %d %d %d %d %d %d ",
(int)job->rusage.ru_maxrss,
(int)job->rusage.ru_ixrss,
(int)job->rusage.ru_idrss,
(int)job->rusage.ru_isrss,
(int)job->rusage.ru_minflt,
- (int)job->rusage.ru_majflt);
- printf("%d %d %d %d %d %d %d %d ",
+ (int)job->rusage.ru_majflt,
(int)job->rusage.ru_nswap,
(int)job->rusage.ru_inblock,
(int)job->rusage.ru_oublock,
@@ -1132,13 +1171,14 @@ void do_dump(void)
(int)job->rusage.ru_msgrcv,
(int)job->rusage.ru_nsignals,
(int)job->rusage.ru_nvcsw,
- (int)job->rusage.ru_nivcsw);
- printf("%d %d %.2f %d %d %.2f "
- "%d %d %.2f %.2f %d %.2f\n",
+ (int)job->rusage.ru_nivcsw,
job->sacct.max_vsize,
+ job->sacct.max_rss);
+ /* Data added in Slurm v1.1 */
+ printf("%d %.2f %d %.2f %d %d %.2f "
+ "%.2f %d %.2f %s %s\n",
job->sacct.max_vsize_task,
job->sacct.ave_vsize,
- job->sacct.max_rss,
job->sacct.max_rss_task,
job->sacct.ave_rss,
job->sacct.max_pages,
@@ -1146,7 +1186,9 @@ void do_dump(void)
job->sacct.ave_pages,
job->sacct.min_cpu,
job->sacct.min_cpu_task,
- job->sacct.ave_cpu);
+ job->sacct.ave_cpu,
+ "-",
+ job->nodes);
}
}
list_iterator_destroy(itr);
@@ -1607,8 +1649,7 @@ void do_list(void)
ListIterator itr_step = NULL;
job_rec_t *job = NULL;
step_rec_t *step = NULL;
- float tempf;
-
+
if (params.opt_total)
do_jobsteps = 0;
@@ -1658,14 +1699,10 @@ void do_list(void)
job->sacct.min_cpu = 0;
if(list_count(job->steps)) {
- tempf = job->sacct.ave_cpu/list_count(job->steps);
- job->sacct.ave_cpu = (uint32_t)tempf;
- tempf = job->sacct.ave_rss/list_count(job->steps);
- job->sacct.ave_rss = (uint32_t)tempf;
- tempf = job->sacct.ave_vsize/list_count(job->steps);
- job->sacct.ave_vsize = (uint32_t)tempf;
- tempf = job->sacct.ave_pages/list_count(job->steps);
- job->sacct.ave_pages = (uint32_t)tempf;
+ job->sacct.ave_cpu /= list_count(job->steps);
+ job->sacct.ave_rss /= list_count(job->steps);
+ job->sacct.ave_vsize /= list_count(job->steps);
+ job->sacct.ave_pages /= list_count(job->steps);
}
if (job->show_full) {
diff --git a/src/sacct/process.c b/src/sacct/process.c
index 78f8476f4f8cebb427a1254924a2af202a544c2f..5f56460b50efff4dee5f591455ca98ae9aeab4ba 100644
--- a/src/sacct/process.c
+++ b/src/sacct/process.c
@@ -203,7 +203,7 @@ int _parse_line(char *f[], void **data)
(*step)->sacct.max_pages = atoi(f[F_MAX_PAGES]);
(*step)->sacct.max_pages_task = atoi(f[F_MAX_PAGES_TASK]);
(*step)->sacct.ave_pages = atof(f[F_AVE_PAGES]);
- (*step)->sacct.min_cpu = atoi(f[F_MIN_CPU]);
+ (*step)->sacct.min_cpu = atof(f[F_MIN_CPU]);
(*step)->sacct.min_cpu_task = atoi(f[F_MIN_CPU_TASK]);
(*step)->sacct.ave_cpu = atof(f[F_AVE_CPU]);
(*step)->stepname = xstrdup(f[F_STEPNAME]);
@@ -328,6 +328,9 @@ got_step:
job->elapsed = time(NULL) - job->header.timestamp;
}
/* now aggregate the aggregatable */
+ job->ncpus = MAX(job->ncpus, step->ncpus);
+ if(step->status < JOB_COMPLETE)
+ return;
job->tot_cpu_sec += step->tot_cpu_sec;
job->tot_cpu_usec += step->tot_cpu_usec;
job->rusage.ru_utime.tv_sec += step->rusage.ru_utime.tv_sec;
@@ -360,10 +363,6 @@ got_step:
/* get the max for all the sacct_t struct */
aggregate_sacct(&job->sacct, &step->sacct);
-
- /* job->psize = MAX(job->psize, step->psize); */
-/* job->vsize = MAX(job->vsize, step->vsize); */
- job->ncpus = MAX(job->ncpus, step->ncpus);
}
void process_suspend(char *f[], int lc, int show_full)
diff --git a/src/sacct/sacct_stat.c b/src/sacct/sacct_stat.c
index f5d68bccca33b13edb03270d28461024183b847c..824448a5441d892059003d131a73c0b339dc534e 100644
--- a/src/sacct/sacct_stat.c
+++ b/src/sacct/sacct_stat.c
@@ -239,7 +239,10 @@ int _sacct_query(resource_allocation_response_msg_t *job, uint32_t step_id)
slurm_mutex_destroy(&stat_mutex);
if(step.ntasks) {
tempf = step.sacct.ave_cpu/step.ntasks;
+ tempf /= 100;
step.sacct.ave_cpu = (uint32_t)tempf;
+ tempf = step.sacct.min_cpu/100;
+ step.sacct.min_cpu = (uint32_t)tempf;
tempf = step.sacct.ave_rss/step.ntasks;
step.sacct.ave_rss = (uint32_t)tempf;
tempf = step.sacct.ave_vsize/step.ntasks;
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index 2df1a4eb823a171c351cb0ebad14425fa8e3f3c6..4d30065250b9e91935f8b722018cb14b381a8e2a 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -792,9 +792,9 @@ void set_slurmd_addr (void)
*/
int update_node ( update_node_msg_t * update_node_msg )
{
- int error_code = 0, base_state, node_inx;
- struct node_record *node_ptr;
- char *this_node_name ;
+ int error_code = 0, base_state = 0, node_inx;
+ struct node_record *node_ptr = NULL;
+ char *this_node_name = NULL;
hostlist_t host_list;
uint16_t node_flags = 0, state_val;
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index 150ec5e378a8489fea6d29933167e65f8428a9c5..e9f7673c11114c1892a3755a34bb2615d735ea99 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -37,6 +37,7 @@
#include <stdlib.h>
#include <sys/types.h>
#include <string.h>
+#include <strings.h>
#include <unistd.h>
#include <slurm/slurm_errno.h>
diff --git a/src/slurmd/slurmd/req.c b/src/slurmd/slurmd/req.c
index 920240d5c91d43e2b6c8641296bbcf71748eec32..e4677e8052093b75f307fae40cc3de12b71f2a6f 100644
--- a/src/slurmd/slurmd/req.c
+++ b/src/slurmd/slurmd/req.c
@@ -1141,7 +1141,8 @@ done2:
close(fd);
done:
slurm_send_rc_msg(msg, rc);
- return SLURM_SUCCESS;
+
+ return rc;
}
static int
diff --git a/src/slurmd/slurmd/slurmd.c b/src/slurmd/slurmd/slurmd.c
index daeee2a856814743ebc3850ec9647933503a4a9e..6bd27b1302ee236fbbeda36068c53267dc3971ea 100644
--- a/src/slurmd/slurmd/slurmd.c
+++ b/src/slurmd/slurmd/slurmd.c
@@ -336,7 +336,6 @@ _handle_connection(slurm_fd fd, slurm_addr *cli)
_service_connection((void *) arg);
return;
}
- slurm_attr_destroy(&attr);
return;
}
diff --git a/src/slurmd/slurmstepd/mgr.c b/src/slurmd/slurmstepd/mgr.c
index 7084c3bd034522dd7be324e135485f0eeb66e4bd..de80f6242eedebb22e2a9675b6cf2da18405d0b0 100644
--- a/src/slurmd/slurmstepd/mgr.c
+++ b/src/slurmd/slurmstepd/mgr.c
@@ -37,6 +37,14 @@
# include <sys/prctl.h>
#endif
+#ifdef HAVE_AIX
+# undef HAVE_UNSETENV
+# include <sys/checkpnt.h>
+#endif
+#ifndef HAVE_UNSETENV
+# include "src/common/unsetenv.h"
+#endif
+
#include <sys/wait.h>
#include <sys/stat.h>
#include <sys/param.h>