From 4535783ddd91bbe7c4944a55c549b6f1947a87c2 Mon Sep 17 00:00:00 2001
From: Morris Jette <jette@schedmd.com>
Date: Mon, 28 Oct 2013 14:53:27 -0700
Subject: [PATCH] Add job array depenency support
Add support for dependencies of job array elements (e.g.
"sbatch --depend=afterok:123_4 ..."). This does not support
depenendencies of ALL job array elements, only individual job
array elements.
---
NEWS | 2 ++
doc/html/job_array.shtml | 9 ++++++++-
src/slurmctld/job_scheduler.c | 33 +++++++++++++++++++++++++++------
3 files changed, 37 insertions(+), 7 deletions(-)
diff --git a/NEWS b/NEWS
index 0bd64ffb4e0..5d9f577cc1d 100644
--- a/NEWS
+++ b/NEWS
@@ -51,6 +51,8 @@ documents those changes that are of interest to users and admins.
-- Fix sacctmgr modify qos set preempt+/-=.
-- BLUEGENE - fix issue where node count wasn't set up correctly when srun
preforms the allocation, regression in 2.6.3.
+ -- Add support for dependencies of job array elements (e.g.
+ "sbatch --depend=afterok:123_4 ...")
* Changes in Slurm 2.6.3
========================
diff --git a/doc/html/job_array.shtml b/doc/html/job_array.shtml
index 326b9b81cf7..05d26e6b650 100644
--- a/doc/html/job_array.shtml
+++ b/doc/html/job_array.shtml
@@ -202,6 +202,13 @@ $ scontrol release 13_2
<h2>Other Command Use</h2>
+<p>Job dependencies for individual job array elements are supported in Slurm
+version 2.6.4 and later. Examples of use follow:</p>
+<pre>
+sbatch --depend=after:123_4 my.job
+sbatch --depend=afterok:123_4:123_8 my.job2
+</pre>
+
<p>The following Slurm commands do not currently recognize job arrays and their
use requires the use of Slurm job IDs, which are unique for each array element:
sacct, sbcast, smap, sreport, sshare, sstat, strigger, and sview.
@@ -245,6 +252,6 @@ array data structure is added rather than the current logic that only
adds a new field to the existing job data structure.
It is not certain when that work will occur.</p>
-<p style="text-align:center;">Last modified 17 September 2013</p>
+<p style="text-align:center;">Last modified 28 October 2013</p>
<!--#include virtual="footer.txt"-->
diff --git a/src/slurmctld/job_scheduler.c b/src/slurmctld/job_scheduler.c
index d4db7b071d6..66dfc07d1e3 100644
--- a/src/slurmctld/job_scheduler.c
+++ b/src/slurmctld/job_scheduler.c
@@ -1527,6 +1527,7 @@ extern int update_job_dependency(struct job_record *job_ptr, char *new_depend)
int rc = SLURM_SUCCESS;
uint16_t depend_type = 0;
uint32_t job_id = 0;
+ uint16_t array_task_id;
char *tok = new_depend, *sep_ptr, *sep_ptr2 = NULL;
List new_depend_list = NULL;
struct depend_spec *dep_ptr;
@@ -1574,21 +1575,33 @@ extern int update_job_dependency(struct job_record *job_ptr, char *new_depend)
sep_ptr = strchr(tok, ':');
if ((sep_ptr == NULL) && (job_id == 0)) {
job_id = strtol(tok, &sep_ptr, 10);
+ if ((sep_ptr != NULL) && (sep_ptr[0] == '_'))
+ array_task_id = strtol(sep_ptr+1, &sep_ptr, 10);
+ else
+ array_task_id = (uint16_t) NO_VAL;
if ((sep_ptr == NULL) || (sep_ptr[0] != '\0') ||
(job_id == 0) || (job_id == job_ptr->job_id)) {
rc = ESLURM_DEPENDENCY;
break;
}
/* old format, just a single job_id */
- dep_job_ptr = find_job_record(job_id);
+ if (array_task_id == (uint16_t) NO_VAL) {
+ dep_job_ptr = find_job_record(job_id);
+ snprintf(dep_buf, sizeof(dep_buf),
+ "afterany:%u", job_id);
+ } else {
+ dep_job_ptr = find_job_array_rec(job_id,
+ array_task_id);
+ snprintf(dep_buf, sizeof(dep_buf),
+ "afterany:%u_%u", job_id,
+ array_task_id);
+ }
if (!dep_job_ptr) /* assume already done */
break;
- snprintf(dep_buf, sizeof(dep_buf),
- "afterany:%u", job_id);
new_depend = dep_buf;
dep_ptr = xmalloc(sizeof(struct depend_spec));
dep_ptr->depend_type = SLURM_DEPEND_AFTER_ANY;
- dep_ptr->job_id = job_id;
+ dep_ptr->job_id = dep_job_ptr->job_id;
dep_ptr->job_ptr = dep_job_ptr;
(void) list_append(new_depend_list, dep_ptr);
break;
@@ -1618,6 +1631,10 @@ extern int update_job_dependency(struct job_record *job_ptr, char *new_depend)
sep_ptr++; /* skip over ":" */
while (rc == SLURM_SUCCESS) {
job_id = strtol(sep_ptr, &sep_ptr2, 10);
+ if ((sep_ptr2 != NULL) && (sep_ptr2[0] == '_'))
+ array_task_id = strtol(sep_ptr2+1,&sep_ptr2,10);
+ else
+ array_task_id = (uint16_t) NO_VAL;
if ((sep_ptr2 == NULL) ||
(job_id == 0) || (job_id == job_ptr->job_id) ||
((sep_ptr2[0] != '\0') && (sep_ptr2[0] != ',') &&
@@ -1625,7 +1642,11 @@ extern int update_job_dependency(struct job_record *job_ptr, char *new_depend)
rc = ESLURM_DEPENDENCY;
break;
}
- dep_job_ptr = find_job_record(job_id);
+ if (array_task_id == (uint16_t) NO_VAL)
+ dep_job_ptr = find_job_record(job_id);
+ else
+ dep_job_ptr = find_job_array_rec(job_id,
+ array_task_id);
if ((depend_type == SLURM_DEPEND_EXPAND) &&
((expand_cnt++ > 0) || (dep_job_ptr == NULL) ||
(!IS_JOB_RUNNING(dep_job_ptr)) ||
@@ -1652,7 +1673,7 @@ extern int update_job_dependency(struct job_record *job_ptr, char *new_depend)
if (dep_job_ptr) { /* job still active */
dep_ptr = xmalloc(sizeof(struct depend_spec));
dep_ptr->depend_type = depend_type;
- dep_ptr->job_id = job_id;
+ dep_ptr->job_id = dep_job_ptr->job_id;
dep_ptr->job_ptr = dep_job_ptr;
(void) list_append(new_depend_list, dep_ptr);
}
--
GitLab