From 0857ad1b77236a4585fea8261dfa9adde0f7376c Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Wed, 11 Feb 2009 18:03:46 +0000
Subject: [PATCH] Add DebugFlag configuration option of "CPU_Bind" for detailed
CPU binding information to be logged.
---
NEWS | 2 ++
doc/man/man5/slurm.conf.5 | 7 +++++--
slurm/slurm.h.in | 2 +-
src/api/config_info.c | 8 ++++----
src/common/read_config.c | 9 ++++++++-
src/common/select_job_res.c | 7 ++++---
src/common/select_job_res.h | 3 ++-
src/plugins/select/cons_res/select_cons_res.c | 12 +++++++++++-
8 files changed, 37 insertions(+), 13 deletions(-)
diff --git a/NEWS b/NEWS
index affccfed6fe..40f3edb3082 100644
--- a/NEWS
+++ b/NEWS
@@ -16,6 +16,8 @@ documents those changes that are of interest to users and admins.
Andersson, National Supercomputer Centre, Sweden.
-- Add support for the scontrol command to be able change the Weight
associated with nodes. Patch from Krishnakumar Ravi[KK] (HP).
+ -- Add DebugFlag configuration option of "CPU_Bind" for detailed CPU
+ binding information to be logged.
* Changes in SLURM 1.4.0-pre7
=============================
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index 42d717a3fb4..8d20987265d 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -240,11 +240,14 @@ Multiple subsystems can be specified with comma separators.
Valid subsystems available today (with more to come) include:
.RS
.TP 15
-\fBTriggers\fR
-Slurmctld triggers
+\fBCPU_Bind\fR
+CPU binding details for jobs and steps
.TP
\fBSteps\fR
Slurmctld resource allocation for job steps
+.TP
+\fBTriggers\fR
+Slurmctld triggers
.RE
.TP
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index 30227885684..6581b888a55 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -1024,7 +1024,7 @@ typedef struct reservation_name_msg {
#define DEBUG_FLAG_SELECT_TYPE 0x00000001 /* SelectType plugin */
#define DEBUG_FLAG_STEPS 0x00000002 /* slurmctld steps */
#define DEBUG_FLAG_TRIGGERS 0x00000004 /* slurmctld triggers */
-
+#define DEBUG_FLAG_CPU_BIND 0x00000008 /* CPU binding */
typedef struct slurm_ctl_conf {
time_t last_update; /* last update time of the build parameters */
diff --git a/src/api/config_info.c b/src/api/config_info.c
index 21ddeae1aae..cd4255d9bd1 100644
--- a/src/api/config_info.c
+++ b/src/api/config_info.c
@@ -148,6 +148,10 @@ void slurm_print_ctl_conf ( FILE* out,
fprintf(out, "CryptoType = %s\n",
slurm_ctl_conf_ptr->crypto_type);
+ xbuf = debug_flags2str(slurm_ctl_conf_ptr->debug_flags);
+ fprintf(out, "DebugFlags = %s\n", xbuf);
+ xfree(xbuf);
+
if (slurm_ctl_conf_ptr->def_mem_per_task & MEM_PER_CPU) {
fprintf(out, "DefMemPerCPU = %u MB\n",
slurm_ctl_conf_ptr->def_mem_per_task &
@@ -158,10 +162,6 @@ void slurm_print_ctl_conf ( FILE* out,
} else
fprintf(out, "DefMemPerCPU = UNLIMITED\n");
- xbuf = debug_flags2str(slurm_ctl_conf_ptr->debug_flags);
- fprintf(out, "DebugFlags = %s\n", xbuf);
- xfree(xbuf);
-
if (slurm_ctl_conf_ptr->disable_root_jobs)
fprintf(out, "DisableRootJobs = YES\n");
else
diff --git a/src/common/read_config.c b/src/common/read_config.c
index beb52b6fbe2..f9d744f1736 100644
--- a/src/common/read_config.c
+++ b/src/common/read_config.c
@@ -2265,6 +2265,11 @@ extern char * debug_flags2str(uint32_t debug_flags)
{
char *rc = NULL;
+ if (debug_flags & DEBUG_FLAG_CPU_BIND) {
+ if (rc)
+ xstrcat(rc, ",");
+ xstrcat(rc, "CPU_Bind");
+ }
if (debug_flags & DEBUG_FLAG_SELECT_TYPE) {
if (rc)
xstrcat(rc, ",");
@@ -2299,7 +2304,9 @@ extern uint32_t debug_str2flags(char *debug_flags)
tmp_str = xstrdup(debug_flags);
tok = strtok_r(tmp_str, ",", &last);
while (tok) {
- if (strcasecmp(tok, "SelectType") == 0)
+ if (strcasecmp(tok, "CPU_Bind") == 0)
+ rc |= DEBUG_FLAG_CPU_BIND;
+ else if (strcasecmp(tok, "SelectType") == 0)
rc |= DEBUG_FLAG_SELECT_TYPE;
else if (strcasecmp(tok, "Steps") == 0)
rc |= DEBUG_FLAG_STEPS;
diff --git a/src/common/select_job_res.c b/src/common/select_job_res.c
index e2f13715bf0..0c988a27113 100644
--- a/src/common/select_job_res.c
+++ b/src/common/select_job_res.c
@@ -409,7 +409,8 @@ extern void free_select_job_res(select_job_res_t *select_job_res_pptr)
}
/* Log the contents of a select_job_res data structure using info() */
-extern void log_select_job_res(select_job_res_t select_job_res_ptr)
+extern void log_select_job_res(uint32_t job_id,
+ select_job_res_t select_job_res_ptr)
{
int bit_inx = 0, bit_reps, i;
int array_size, node_inx;
@@ -421,8 +422,8 @@ extern void log_select_job_res(select_job_res_t select_job_res_ptr)
}
info("====================");
- info("nhosts:%u nprocs:%u node_req:%u",
- select_job_res_ptr->nhosts, select_job_res_ptr->nprocs,
+ info("job_id:%u nhosts:%u nprocs:%u node_req:%u",
+ job_id, select_job_res_ptr->nhosts, select_job_res_ptr->nprocs,
select_job_res_ptr->node_req);
if (select_job_res_ptr->cpus == NULL) {
diff --git a/src/common/select_job_res.h b/src/common/select_job_res.h
index aa462c644c8..719ad16a2b0 100644
--- a/src/common/select_job_res.h
+++ b/src/common/select_job_res.h
@@ -170,7 +170,8 @@ extern select_job_res_t copy_select_job_res(select_job_res_t
extern void free_select_job_res(select_job_res_t *select_job_res_pptr);
/* Log the contents of a select_job_res data structure using info() */
-extern void log_select_job_res(select_job_res_t select_job_res_ptr);
+extern void log_select_job_res(uint32_t job_id,
+ select_job_res_t select_job_res_ptr);
/* Un/pack full select_job_res data structure */
extern void pack_select_job_res(select_job_res_t select_job_res_ptr,
diff --git a/src/plugins/select/cons_res/select_cons_res.c b/src/plugins/select/cons_res/select_cons_res.c
index a954440eaae..f590832f0cf 100644
--- a/src/plugins/select/cons_res/select_cons_res.c
+++ b/src/plugins/select/cons_res/select_cons_res.c
@@ -1207,9 +1207,16 @@ extern int select_p_job_test(struct job_record *job_ptr, bitstr_t * bitmap,
{
int rc;
uint16_t job_node_req;
+ bool debug_cpu_bind = false, debug_check = false;
xassert(bitmap);
+ if (!debug_check) {
+ debug_check = true;
+ if (slurm_get_debug_flags() & DEBUG_FLAG_CPU_BIND)
+ debug_cpu_bind = true;
+ }
+
if (!job_ptr->details)
return EINVAL;
@@ -1237,10 +1244,13 @@ extern int select_p_job_test(struct job_record *job_ptr, bitstr_t * bitmap,
#if (CR_DEBUG)
if (job_ptr->select_job)
- log_select_job_res(job_ptr->select_job);
+ log_select_job_res(job_ptr->job_id, job_ptr->select_job);
else
info("no select_job_res info for job %u",
job_ptr->job_id);
+#else
+ if (debug_cpu_bind)
+ log_select_job_res(job_ptr->job_id, job_ptr->select_job);
#endif
return rc;
--
GitLab