diff --git a/NEWS b/NEWS
index 5d6e1ca46aa42c1d2c05bc9d4401916d43270618..97f7ce17b1de75dbd014e6823fbba99dfbf7a5cf 100644
--- a/NEWS
+++ b/NEWS
@@ -95,6 +95,9 @@ documents those changes that are of interest to users and administrators.
==========================
-- Docs - the correct default value for GroupUpdateForce is 0.
-- mpi/pmix - improve point to point communication performance.
+ -- SlurmDB - include pending jobs in search during 'sacctmgr show runawayjobs'.
+ -- Add client side out-of-range checks to --nice flag.
+ -- Fix support for sbatch "-W" option, previously eeded to use "--wait".
* Changes in Slurm 16.05.5
==========================
diff --git a/src/plugins/accounting_storage/mysql/as_mysql_fix_runaway_jobs.c b/src/plugins/accounting_storage/mysql/as_mysql_fix_runaway_jobs.c
index 6e06aed3a319f404268e5bcb1424153e90c15874..89ca7e484a446a5716a94b9f105813e8cccc5ba4 100644
--- a/src/plugins/accounting_storage/mysql/as_mysql_fix_runaway_jobs.c
+++ b/src/plugins/accounting_storage/mysql/as_mysql_fix_runaway_jobs.c
@@ -120,7 +120,8 @@ extern int as_mysql_fix_runaway_jobs(mysql_conn_t *mysql_conn, uint32_t uid,
xstrfmtcat(job_ids, "%s%d", ((job_ids) ? "," : ""), job->jobid);
}
- query = xstrdup_printf("UPDATE \"%s_%s\" SET time_end=time_start, "
+ query = xstrdup_printf("UPDATE \"%s_%s\" SET time_end="
+ "GREATEST(time_start, time_eligible, time_submit), "
"state=%d WHERE id_job IN (%s);",
mysql_conn->cluster_name, job_table,
JOB_COMPLETE, job_ids);
diff --git a/src/sacctmgr/runaway_job_functions.c b/src/sacctmgr/runaway_job_functions.c
index f1dbff24e47b33ceeb050883475c74503e54add2..43916592de0885ea87a4ea6fc54a84897e1d6613 100644
--- a/src/sacctmgr/runaway_job_functions.c
+++ b/src/sacctmgr/runaway_job_functions.c
@@ -200,6 +200,7 @@ static List _get_runaway_jobs(slurmdb_job_cond_t *job_cond)
job_cond->without_steps = 1;
job_cond->without_usage_truncation = 1;
job_cond->state_list = list_create(slurm_destroy_char);
+ slurm_addto_char_list(job_cond->state_list, "0");
slurm_addto_char_list(job_cond->state_list, "1");
if (!job_cond->cluster_list || !list_count(job_cond->cluster_list)) {
@@ -290,9 +291,11 @@ extern int sacctmgr_list_runaway_jobs(int argc, char *argv[])
List format_list = list_create(slurm_destroy_char);
slurmdb_job_cond_t *job_cond = xmalloc(sizeof(slurmdb_job_cond_t));
char *ask_msg = "\nWould you like to fix these runaway jobs?\n"
- "(This will set the end times to start times and "
- "states to completed for these jobs and will trigger "
- "the rollup to reroll usage from before the oldest "
+ "(This will set the end time for each job to the "
+ "latest out of the start, eligible, or submit times, "
+ "and set the state to completed.\n"
+ "Once corrected, this will trigger the rollup to "
+ "reroll usage from before the oldest "
"runaway job.)\n\n";
diff --git a/src/salloc/opt.c b/src/salloc/opt.c
index 9862d9c7f7a776ca0f15cd5f006aaf3ff3363ecb..c57164812bf72718771dad8ed7ab43036289dcd8 100644
--- a/src/salloc/opt.c
+++ b/src/salloc/opt.c
@@ -1101,21 +1101,29 @@ void set_options(const int argc, char **argv)
opt.mcs_label = xstrdup(optarg);
break;
}
- case LONG_OPT_NICE:
+ case LONG_OPT_NICE: {
+ long long tmp_nice;
if (optarg)
- opt.nice = strtol(optarg, NULL, 10);
+ tmp_nice = strtoll(optarg, NULL, 10);
else
- opt.nice = 100;
- if (opt.nice < 0) {
+ tmp_nice = 100;
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
+ tmp_nice = 0;
+ }
+ if (tmp_nice < 0) {
uid_t my_uid = getuid();
if ((my_uid != 0) &&
(my_uid != slurm_get_slurm_user_id())) {
error("Nice value must be "
"non-negative, value ignored");
- opt.nice = 0;
+ tmp_nice = 0;
}
}
+ opt.nice = (int) tmp_nice;
break;
+ }
case LONG_OPT_PRIORITY:
if (!optarg) { /* CLANG Fix */
;
diff --git a/src/sbatch/opt.c b/src/sbatch/opt.c
index fb7a2653dc27433f8e07519bc2da5d2bc0be4cfc..f1eab0778331c52901ca591b153999a0cb595263 100644
--- a/src/sbatch/opt.c
+++ b/src/sbatch/opt.c
@@ -124,7 +124,6 @@ enum wrappers {
#define LONG_OPT_PROPAGATE 0x100
#define LONG_OPT_MEM_BIND 0x102
#define LONG_OPT_POWER 0x103
-#define LONG_OPT_WAIT 0x104
#define LONG_OPT_JOBID 0x105
#define LONG_OPT_TMP 0x106
#define LONG_OPT_MEM 0x107
@@ -807,6 +806,7 @@ static struct option long_options[] = {
{"verbose", no_argument, 0, 'v'},
{"version", no_argument, 0, 'V'},
{"nodelist", required_argument, 0, 'w'},
+ {"wait", no_argument, 0, 'W'},
{"exclude", required_argument, 0, 'x'},
{"acctg-freq", required_argument, 0, LONG_OPT_ACCTG_FREQ},
{"bbf", required_argument, 0, LONG_OPT_BURST_BUFFER_FILE},
@@ -874,7 +874,6 @@ static struct option long_options[] = {
{"tmp", required_argument, 0, LONG_OPT_TMP},
{"uid", required_argument, 0, LONG_OPT_UID},
{"use-min-nodes", no_argument, 0, LONG_OPT_USE_MIN_NODES},
- {"wait", no_argument, 0, LONG_OPT_WAIT},
{"wait-all-nodes",required_argument, 0, LONG_OPT_WAIT_ALL_NODES},
{"wckey", required_argument, 0, LONG_OPT_WCKEY},
{"wrap", required_argument, 0, LONG_OPT_WRAP},
@@ -1509,6 +1508,9 @@ static void _set_options(int argc, char **argv)
xfree(opt.nodelist);
opt.nodelist = xstrdup(optarg);
break;
+ case 'W':
+ opt.wait = true;
+ break;
case 'x':
xfree(opt.exc_nodes);
opt.exc_nodes = xstrdup(optarg);
@@ -1669,21 +1671,29 @@ static void _set_options(int argc, char **argv)
xfree(opt.burst_buffer_file);
opt.burst_buffer_file = _read_file(optarg);
break;
- case LONG_OPT_NICE:
+ case LONG_OPT_NICE: {
+ long long tmp_nice;
if (optarg)
- opt.nice = strtol(optarg, NULL, 10);
+ tmp_nice = strtoll(optarg, NULL, 10);
else
- opt.nice = 100;
- if (opt.nice < 0) {
+ tmp_nice = 100;
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
+ tmp_nice = 0;
+ }
+ if (tmp_nice < 0) {
uid_t my_uid = getuid();
if ((my_uid != 0) &&
(my_uid != slurm_get_slurm_user_id())) {
error("Nice value must be "
"non-negative, value ignored");
- opt.nice = 0;
+ tmp_nice = 0;
}
}
+ opt.nice = (int) tmp_nice;
break;
+ }
case LONG_OPT_PRIORITY:
if (!optarg) { /* CLANG Fix */
;
@@ -1839,9 +1849,6 @@ static void _set_options(int argc, char **argv)
xfree(opt.network);
opt.network = xstrdup(optarg);
break;
- case LONG_OPT_WAIT:
- opt.wait = true;
- break;
case LONG_OPT_WCKEY:
xfree(opt.wckey);
opt.wckey = xstrdup(optarg);
@@ -2202,21 +2209,29 @@ static void _set_pbs_options(int argc, char **argv)
else
opt.ofname = xstrdup(optarg);
break;
- case 'p':
+ case 'p': {
+ long long tmp_nice;
if (optarg)
- opt.nice = strtol(optarg, NULL, 10);
+ tmp_nice = strtoll(optarg, NULL, 10);
else
- opt.nice = 100;
- if (opt.nice < 0) {
+ tmp_nice = 100;
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
+ tmp_nice = 0;
+ }
+ if (tmp_nice < 0) {
uid_t my_uid = getuid();
if ((my_uid != 0) &&
(my_uid != slurm_get_slurm_user_id())) {
error("Nice value must be "
"non-negative, value ignored");
- opt.nice = 0;
+ tmp_nice = 0;
}
}
+ opt.nice = (int) tmp_nice;
break;
+ }
case 'q':
xfree(opt.partition);
opt.partition = xstrdup(optarg);
@@ -2515,21 +2530,28 @@ static void _parse_pbs_resource_list(char *rl)
xfree(temp);
}
} else if (!xstrncmp(rl+i, "nice=", 5)) {
+ long long tmp_nice;
i += 5;
temp = _get_pbs_option_value(rl, &i, ',');
if (temp)
- opt.nice = strtol(temp, NULL, 10);
+ tmp_nice = strtoll(temp, NULL, 10);
else
- opt.nice = 100;
- if (opt.nice < 0) {
+ tmp_nice = 100;
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
+ tmp_nice = 0;
+ }
+ if (tmp_nice < 0) {
uid_t my_uid = getuid();
if ((my_uid != 0) &&
(my_uid != slurm_get_slurm_user_id())) {
error("Nice value must be "
"non-negative, value ignored");
- opt.nice = 0;
+ tmp_nice = 0;
}
}
+ opt.nice = (int) tmp_nice;
xfree(temp);
} else if (!xstrncmp(rl+i, "nodes=", 6)) {
i+=6;
diff --git a/src/scontrol/update_job.c b/src/scontrol/update_job.c
index 7724ba543c62fbaba529a385b2546fe18ae9529a..b433afec1b5e30e9ecef19029856c539e4405751 100644
--- a/src/scontrol/update_job.c
+++ b/src/scontrol/update_job.c
@@ -856,16 +856,15 @@ scontrol_update_job (int argc, char *argv[])
update_cnt++;
}
else if (strncasecmp(tag, "Nice", MAX(taglen, 2)) == 0) {
- int nice;
- nice = strtoll(val, (char **) NULL, 10);
- if (abs(nice) > NICE_OFFSET) {
- error("Invalid nice value, must be between "
- "-%d and %d", NICE_OFFSET,
- NICE_OFFSET);
+ long long tmp_nice;
+ tmp_nice = strtoll(val, (char **)NULL, 10);
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
exit_code = 1;
return 0;
}
- job_msg.nice = NICE_OFFSET + nice;
+ job_msg.nice = NICE_OFFSET + tmp_nice;
update_cnt++;
}
else if (strncasecmp(tag, "CPUsPerTask", MAX(taglen, 6)) == 0) {
diff --git a/src/srun/libsrun/opt.c b/src/srun/libsrun/opt.c
index 0b4370bf60e31a59734496aefffa9f6aaf922c0d..684644fe8798174060ab55b758c16d8014647aa9 100644
--- a/src/srun/libsrun/opt.c
+++ b/src/srun/libsrun/opt.c
@@ -1551,21 +1551,29 @@ static void _set_options(const int argc, char **argv)
xfree(opt.task_epilog);
opt.task_epilog = xstrdup(optarg);
break;
- case LONG_OPT_NICE:
+ case LONG_OPT_NICE: {
+ long long tmp_nice;
if (optarg)
- opt.nice = strtol(optarg, NULL, 10);
+ tmp_nice = strtoll(optarg, NULL, 10);
else
- opt.nice = 100;
- if (opt.nice < 0) {
+ tmp_nice = 100;
+ if (abs(tmp_nice) > (NICE_OFFSET - 3)) {
+ error("Nice value out of range (+/- %u). Value "
+ "ignored", NICE_OFFSET - 3);
+ tmp_nice = 0;
+ }
+ if (tmp_nice < 0) {
uid_t my_uid = getuid();
if ((my_uid != 0) &&
(my_uid != slurm_get_slurm_user_id())) {
- error("Nice value must be non-negative, "
- "value ignored");
- opt.nice = 0;
+ error("Nice value must be "
+ "non-negative, value ignored");
+ tmp_nice = 0;
}
}
+ opt.nice = (int) tmp_nice;
break;
+ }
case LONG_OPT_PRIORITY: {
long long priority;
if (strcasecmp(optarg, "TOP") == 0) {