From c45b5c4689a05a862405d3dafca60221d0bb723a Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Thu, 25 Aug 2005 19:23:53 +0000
Subject: [PATCH] Add name and network parameters to job step information,
reported by squeue and scontrol, data saved/restored by slurmctld, etc.
---
NEWS | 2 ++
doc/man/man1/squeue.1 | 18 ++++++++--------
slurm/slurm.h.in | 4 ++++
src/api/job_step_info.c | 5 +++--
src/common/slurm_protocol_defs.c | 2 ++
src/common/slurm_protocol_pack.c | 19 ++++++++++++++---
src/common/slurm_protocol_pack.h | 3 ++-
src/slurmctld/job_mgr.c | 11 +++++++++-
src/slurmctld/proc_req.c | 5 ++++-
src/slurmctld/slurmctld.h | 7 ++++---
src/slurmctld/step_mgr.c | 35 +++++++++++++++++++++++---------
src/squeue/opts.c | 6 ++++++
src/squeue/print.c | 12 +++++++++++
src/squeue/print.h | 4 ++++
src/squeue/squeue.c | 2 +-
src/srun/allocate.c | 2 ++
16 files changed, 106 insertions(+), 31 deletions(-)
diff --git a/NEWS b/NEWS
index e34b78336ef..489f0badb4d 100644
--- a/NEWS
+++ b/NEWS
@@ -5,6 +5,8 @@ documents those changes that are of interest to users and admins.
=============================
-- Add code so job request for shared nodes gets explicitly requested
nodes, but lightly loaded nodes otherwise.
+ -- Add job step name field.
+ -- Add job step network specification field.
-- Add proctrack/rms plugin
-- Change the proctrack API to send a slurmd_job_t pointer to both
slurm_container_create() and slurm_container_add(). One of those
diff --git a/doc/man/man1/squeue.1 b/doc/man/man1/squeue.1
index 34195c5ada2..0afbf02c22c 100644
--- a/doc/man/man1/squeue.1
+++ b/doc/man/man1/squeue.1
@@ -1,4 +1,4 @@
-.TH SQUEUE "1" "July 2005" "squeue 0.5" "Slurm components"
+.TH SQUEUE "1" "August 2005" "squeue 0.6" "Slurm components"
.SH "NAME"
squeue \- view information about jobs located in the SLURM scheduling queue.
@@ -62,7 +62,7 @@ various options are
".7i %.9P %.8j %.8u %.8T %.9M %.9l %.6D %R"
.TP
.I "-s, --steps"
-"%10i %.9P %.8u %.9M %N"
+"%10i %.8j %.9P %.8u %.9M %N"
.RE
.IP
@@ -118,7 +118,7 @@ Can the nodes allocated to the job be shared with other jobs
Job or job step id
.TP
\fB%j\fR
-Job name
+Job or job step name
.TP
\fB%l\fR
Time limit of the job in days:hours:minutes:seconds.
@@ -355,13 +355,13 @@ Print the job steps in the debug partition sorted by user:
.br
# squeue -s -p debug -S u
.br
- STEPID PARTITION USER TIME_USED NODELIST(REASON)
+ STEPID NAME PARTITION USER TIME_USED NODELIST(REASON)
.br
- 65552.1 debug alice 0:23 dev[1-4]
+ 65552.1 test1 debug alice 0:23 dev[1-4]
.br
- 65562.2 debug bob 0:18 dev22
+ 65562.2 big_run debug bob 0:18 dev22
.br
- 65550.1 debug candice 1:43:21 dev[6-12]
+ 65550.1 param1 debug candice 1:43:21 dev[6-12]
.ec
.eo
@@ -383,9 +383,9 @@ Print information only about job step 65552.1:
.br
# squeue --steps 65552.1
.br
- STEPID PARTITION USER TIME_USED NODELIST(REASON)
+ STEPID NAME PARTITION USER TIME_USED NODELIST(REASON)
.br
- 65552.1 debug alice 12:49 dev[1-4]
+ 65552.1 test2 debug alice 12:49 dev[1-4]
.ec
.SH "COPYING"
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index 14700786571..db0f0bcfcc0 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -379,6 +379,8 @@ typedef struct job_step_specs {
uint16_t port; /* port to contact initiating srun */
char *host; /* host to contact initiating srun */
char *node_list; /* list of required nodes */
+ char *network; /* network use spec */
+ char *name; /* name of the job step, default "" */
} job_step_create_request_msg_t;
typedef struct job_step_create_response_msg {
@@ -396,6 +398,8 @@ typedef struct {
time_t start_time; /* step start time */
char *partition; /* name of assigned partition */
char *nodes; /* list of nodes allocated to job_step */
+ char *name; /* name of job step */
+ char *network; /* network specs for job step */
} job_step_info_t;
typedef struct job_step_info_response_msg {
diff --git a/src/api/job_step_info.c b/src/api/job_step_info.c
index b1905bf7559..04ab48a3231 100644
--- a/src/api/job_step_info.c
+++ b/src/api/job_step_info.c
@@ -91,8 +91,9 @@ slurm_print_job_step_info ( FILE* out, job_step_info_t * job_step_ptr,
fprintf ( out, "\n ");
/****** Line 2 ******/
- fprintf ( out, "Partition=%s Nodes=%s\n\n",
- job_step_ptr->partition, job_step_ptr->nodes);
+ fprintf ( out, "Partition=%s Nodes=%s Name=%s Network=%s\n\n",
+ job_step_ptr->partition, job_step_ptr->nodes,
+ job_step_ptr->name, job_step_ptr->network);
}
/*
diff --git a/src/common/slurm_protocol_defs.c b/src/common/slurm_protocol_defs.c
index f4de323ad19..212c79f3eb3 100644
--- a/src/common/slurm_protocol_defs.c
+++ b/src/common/slurm_protocol_defs.c
@@ -254,6 +254,8 @@ void slurm_free_job_step_create_request_msg(job_step_create_request_msg_t *
msg)
{
if (msg) {
+ xfree(msg->name);
+ xfree(msg->network);
xfree(msg->node_list);
xfree(msg->host);
xfree(msg);
diff --git a/src/common/slurm_protocol_pack.c b/src/common/slurm_protocol_pack.c
index f4a0fd0bf6d..163a2630216 100644
--- a/src/common/slurm_protocol_pack.c
+++ b/src/common/slurm_protocol_pack.c
@@ -1296,6 +1296,8 @@ _pack_job_step_create_request_msg(job_step_create_request_msg_t
pack16(msg->task_dist, buffer);
pack16(msg->port, buffer);
packstr(msg->host, buffer);
+ packstr(msg->name, buffer);
+ packstr(msg->network, buffer);
packstr(msg->node_list, buffer);
}
@@ -1321,12 +1323,16 @@ _unpack_job_step_create_request_msg(job_step_create_request_msg_t ** msg,
safe_unpack16(&(tmp_ptr->task_dist), buffer);
safe_unpack16(&(tmp_ptr->port), buffer);
safe_unpackstr_xmalloc(&(tmp_ptr->host), &uint16_tmp, buffer);
+ safe_unpackstr_xmalloc(&(tmp_ptr->name), &uint16_tmp, buffer);
+ safe_unpackstr_xmalloc(&(tmp_ptr->network), &uint16_tmp, buffer);
safe_unpackstr_xmalloc(&(tmp_ptr->node_list), &uint16_tmp, buffer);
return SLURM_SUCCESS;
unpack_error:
xfree(tmp_ptr->host);
+ xfree(tmp_ptr->name);
+ xfree(tmp_ptr->network);
xfree(tmp_ptr->node_list);
xfree(tmp_ptr);
*msg = NULL;
@@ -1572,7 +1578,8 @@ void
pack_job_step_info_members(uint32_t job_id, uint16_t step_id,
uint32_t user_id, uint32_t num_tasks,
time_t start_time, char *partition,
- char *nodes, Buf buffer)
+ char *nodes, char *name, char *network,
+ Buf buffer)
{
pack32(job_id, buffer);
pack16(step_id, buffer);
@@ -1582,7 +1589,8 @@ pack_job_step_info_members(uint32_t job_id, uint16_t step_id,
pack_time(start_time, buffer);
packstr(partition, buffer);
packstr(nodes, buffer);
-
+ packstr(name, buffer);
+ packstr(network, buffer);
}
/* pack_job_step_info
@@ -1599,7 +1607,8 @@ pack_job_step_info(job_step_info_t * step, Buf buffer)
step->user_id,
step->num_tasks,
step->start_time,
- step->partition, step->nodes, buffer);
+ step->partition, step->nodes,
+ step->name, step->network, buffer);
}
/* _unpack_job_step_info_members
@@ -1621,12 +1630,16 @@ _unpack_job_step_info_members(job_step_info_t * step, Buf buffer)
safe_unpack_time(&step->start_time, buffer);
safe_unpackstr_xmalloc(&step->partition, &uint16_tmp, buffer);
safe_unpackstr_xmalloc(&step->nodes, &uint16_tmp, buffer);
+ safe_unpackstr_xmalloc(&step->name, &uint16_tmp, buffer);
+ safe_unpackstr_xmalloc(&step->network, &uint16_tmp, buffer);
return SLURM_SUCCESS;
unpack_error:
xfree(step->partition);
xfree(step->nodes);
+ xfree(step->name);
+ xfree(step->network);
return SLURM_ERROR;
}
diff --git a/src/common/slurm_protocol_pack.h b/src/common/slurm_protocol_pack.h
index 3d5cfe58cb0..363abfd302e 100644
--- a/src/common/slurm_protocol_pack.h
+++ b/src/common/slurm_protocol_pack.h
@@ -124,6 +124,7 @@ void pack_job_step_info ( job_step_info_t* step, Buf buffer );
*/
void pack_job_step_info_members( uint32_t job_id, uint16_t step_id,
uint32_t user_id, uint32_t num_tasks, time_t start_time,
- char *partition, char *nodes, Buf buffer );
+ char *partition, char *nodes, char *name, char *network,
+ Buf buffer );
#endif
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 2e4f818c5de..aca736508e2 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -72,7 +72,7 @@
#define JOB_HASH_INX(_job_id) (_job_id % hash_table_size)
-#define JOB_STATE_VERSION "VER001"
+#define JOB_STATE_VERSION "VER002"
/* Global variables */
List job_list = NULL; /* job_record list */
@@ -816,6 +816,8 @@ static void _dump_job_step_state(struct step_record *step_ptr, Buf buffer)
pack_time(step_ptr->start_time, buffer);
packstr(step_ptr->host, buffer);
packstr(step_ptr->step_node_list, buffer);
+ packstr(step_ptr->name, buffer);
+ packstr(step_ptr->network, buffer);
pack16(step_ptr->batch_step, buffer);
if (!step_ptr->batch_step)
switch_pack_jobinfo(step_ptr->switch_job, buffer);
@@ -830,6 +832,7 @@ static int _load_step_state(struct job_record *job_ptr, Buf buffer)
uint32_t num_tasks;
time_t start_time;
char *step_node_list = NULL, *host = NULL;
+ char *name = NULL, *network = NULL;
switch_jobinfo_t switch_tmp = NULL;
check_jobinfo_t check_tmp = NULL;
@@ -840,6 +843,8 @@ static int _load_step_state(struct job_record *job_ptr, Buf buffer)
safe_unpack_time(&start_time, buffer);
safe_unpackstr_xmalloc(&host, &name_len, buffer);
safe_unpackstr_xmalloc(&step_node_list, &name_len, buffer);
+ safe_unpackstr_xmalloc(&name, &name_len, buffer);
+ safe_unpackstr_xmalloc(&network, &name_len, buffer);
safe_unpack16(&batch_step, buffer);
if (!batch_step) {
switch_alloc_jobinfo(&switch_tmp);
@@ -869,6 +874,8 @@ static int _load_step_state(struct job_record *job_ptr, Buf buffer)
/* set new values */
step_ptr->step_id = step_id;
step_ptr->cyclic_alloc = cyclic_alloc;
+ step_ptr->name = name;
+ step_ptr->network = network;
step_ptr->num_tasks = num_tasks;
step_ptr->port = port;
step_ptr->host = host;
@@ -885,6 +892,8 @@ static int _load_step_state(struct job_record *job_ptr, Buf buffer)
unpack_error:
xfree(host);
+ xfree(name);
+ xfree(network);
xfree(step_node_list);
if (switch_tmp) switch_free_jobinfo(switch_tmp);
return SLURM_FAILURE;
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index 8598dbf6b37..067911402b7 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -558,6 +558,8 @@ static void _slurm_rpc_allocate_and_run(slurm_msg_t * msg)
req_step_msg.node_count = INFINITE;
req_step_msg.cpu_count = job_desc_msg->num_procs;
#endif
+ req_step_msg.name = job_ptr->name;
+ req_step_msg.network = job_ptr->network;
req_step_msg.num_tasks = job_desc_msg->num_tasks;
req_step_msg.task_dist = job_desc_msg->task_dist;
error_code = step_create(&req_step_msg, &step_rec, true, false);
@@ -2037,8 +2039,9 @@ int _launch_batch_step(job_desc_msg_t *job_desc_msg, uid_t uid,
req_step_msg.task_dist = SLURM_DIST_CYCLIC;
req_step_msg.port = 0;
req_step_msg.host = NULL;
+ req_step_msg.name = NULL;
+ req_step_msg.network = NULL;
req_step_msg.node_list = NULL;
-
START_TIMER;
lock_slurmctld(job_write_lock);
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index 97ed8ad7225..14307de760a 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -349,9 +349,10 @@ struct step_record {
uint16_t batch_step; /* 1 if batch job step, 0 otherwise */
switch_jobinfo_t switch_job; /* switch context, opaque */
check_jobinfo_t check_job; /* checkpoint context, opaque */
+ char *name; /* name of job step */
+ char *network; /* step's network specification */
};
-typedef struct job_step_specs step_specs;
extern List job_list; /* list of job_record entries */
extern List job_list; /* list of job_record entries */
@@ -506,7 +507,7 @@ extern void dump_job_desc(job_desc_msg_t * job_specs);
* dump_step_desc - dump the incoming step initiate request message
* IN step_spec - job step request specification from RPC
*/
-extern void dump_step_desc(step_specs *step_spec);
+extern void dump_step_desc(job_step_create_request_msg_t *step_spec);
/*
* find_job_record - return a pointer to the job record with the given job_id
@@ -1107,7 +1108,7 @@ extern int slurmctld_shutdown(void);
* NOTE: don't free the returned step_record because that is managed through
* the job.
*/
-extern int step_create ( step_specs *step_specs,
+extern int step_create ( job_step_create_request_msg_t *step_specs,
struct step_record** new_step_record,
bool kill_job_when_step_done,
bool batch_step );
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index ab3558c6f26..701a8e4aa23 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -56,7 +56,7 @@
static int _job_step_ckpt_error(struct step_record *step_ptr, slurm_fd conn_fd);
static void _pack_ctld_job_step_info(struct step_record *step, Buf buffer);
static bitstr_t * _pick_step_nodes (struct job_record *job_ptr,
- step_specs *step_spec );
+ job_step_create_request_msg_t *step_spec );
/*
* create_step_record - create an empty step_record for the specified job.
* IN job_ptr - pointer to job table entry to have step record added
@@ -166,7 +166,7 @@ delete_step_record (struct job_record *job_ptr, uint32_t step_id)
* IN step_spec - job step request specification from RPC
*/
void
-dump_step_desc(step_specs *step_spec)
+dump_step_desc(job_step_create_request_msg_t *step_spec)
{
if (step_spec == NULL)
return;
@@ -177,8 +177,9 @@ dump_step_desc(step_specs *step_spec)
debug3(" num_tasks=%u relative=%u task_dist=%u node_list=%s",
step_spec->num_tasks, step_spec->relative,
step_spec->task_dist, step_spec->node_list);
- debug3(" host=%s port=%u",
- step_spec->host, step_spec->port);
+ debug3(" host=%s port=%u name=%s network=%s",
+ step_spec->host, step_spec->port, step_spec->name,
+ step_spec->network);
}
@@ -366,7 +367,8 @@ int job_step_complete(uint32_t job_id, uint32_t step_id, uid_t uid,
* NOTE: returned bitmap must be freed by the caller using bit_free()
*/
static bitstr_t *
-_pick_step_nodes (struct job_record *job_ptr, step_specs *step_spec )
+_pick_step_nodes (struct job_record *job_ptr,
+ job_step_create_request_msg_t *step_spec )
{
bitstr_t *nodes_avail = NULL, *nodes_picked = NULL, *node_tmp = NULL;
@@ -488,9 +490,10 @@ cleanup:
* NOTE: don't free the returned step_record because that is managed through
* the job.
*/
-int
-step_create ( step_specs *step_specs, struct step_record** new_step_record,
- bool kill_job_when_step_done, bool batch_step )
+extern int
+step_create ( job_step_create_request_msg_t *step_specs,
+ struct step_record** new_step_record,
+ bool kill_job_when_step_done, bool batch_step )
{
struct step_record *step_ptr;
struct job_record *job_ptr;
@@ -554,6 +557,17 @@ step_create ( step_specs *step_specs, struct step_record** new_step_record,
step_ptr->host = xstrdup(step_specs->host);
step_ptr->batch_step = batch_step;
+ /* step's name and network default to job's values if not
+ * specified in the step specification */
+ if (step_specs->name && step_specs->name[0])
+ step_ptr->name = xstrdup(step_specs->name);
+ else
+ step_ptr->name = xstrdup(job_ptr->name);
+ if (step_specs->network && step_specs->network[0])
+ step_ptr->network = xstrdup(step_specs->network);
+ else
+ step_ptr->network = xstrdup(job_ptr->network);
+
/* a batch script does not need switch info */
if (!batch_step) {
int *tasks_per_node;
@@ -571,7 +585,7 @@ step_create ( step_specs *step_specs, struct step_record** new_step_record,
step_ptr->step_node_list,
tasks_per_node,
step_ptr->cyclic_alloc,
- job_ptr->network) < 0) {
+ step_ptr->network) < 0) {
error("switch_build_jobinfo: %m");
xfree(tasks_per_node);
delete_step_record (job_ptr, step_ptr->step_id);
@@ -598,7 +612,8 @@ static void _pack_ctld_job_step_info(struct step_record *step, Buf buffer)
step->num_tasks,
step->start_time,
step->job_ptr->partition,
- step->step_node_list, buffer);
+ step->step_node_list,
+ step->name, step->network, buffer);
}
/*
diff --git a/src/squeue/opts.c b/src/squeue/opts.c
index d5c822f3b21..03c6a5df19a 100644
--- a/src/squeue/opts.c
+++ b/src/squeue/opts.c
@@ -335,6 +335,12 @@ extern int parse_format( char* format )
step_format_add_id( params.format_list,
field_size,
right_justify, suffix );
+ else if (field[0] == 'j')
+ step_format_add_name( params.format_list,
+ field_size,
+ right_justify,
+ suffix );
+
else if (field[0] == 'M')
step_format_add_time_used( params.format_list,
field_size,
diff --git a/src/squeue/print.c b/src/squeue/print.c
index 246bca7ce0b..23c9ee985fb 100644
--- a/src/squeue/print.c
+++ b/src/squeue/print.c
@@ -950,6 +950,18 @@ int _print_step_time_used(job_step_info_t * step, int width, bool right,
return SLURM_SUCCESS;
}
+int _print_step_name(job_step_info_t * step, int width, bool right,
+ char* suffix)
+{
+ if (step == NULL) /* Print the Header instead */
+ _print_str("NAME", width, right, true);
+ else
+ _print_nodes(step->name, width, right, true);
+ if (suffix)
+ printf("%s", suffix);
+ return SLURM_SUCCESS;
+}
+
int _print_step_nodes(job_step_info_t * step, int width, bool right,
char* suffix)
{
diff --git a/src/squeue/print.h b/src/squeue/print.h
index 7c879fddea3..0d34ee54db9 100644
--- a/src/squeue/print.h
+++ b/src/squeue/print.h
@@ -236,6 +236,8 @@ int step_format_add_function(List list, int width, bool right_justify,
step_format_add_function(list,wid,right,suffix,_print_step_time_used)
#define step_format_add_nodes(list,wid,right,suffix) \
step_format_add_function(list,wid,right,suffix,_print_step_nodes)
+#define step_format_add_name(list,wid,right,suffix) \
+ step_format_add_function(list,wid,right,suffix,_print_step_name)
/*****************************************************************************
* Step Line Print Functions
@@ -254,6 +256,8 @@ int _print_step_time_start(job_step_info_t * step, int width,
bool right_justify, char *suffix);
int _print_step_time_used(job_step_info_t * step, int width,
bool right_justify, char *suffix);
+int _print_step_name(job_step_info_t * step, int width,
+ bool right_justify, char *suffix);
int _print_step_nodes(job_step_info_t * step, int width,
bool right_justify, char *suffix);
diff --git a/src/squeue/squeue.c b/src/squeue/squeue.c
index 03b44545d2f..f7d05e94eac 100644
--- a/src/squeue/squeue.c
+++ b/src/squeue/squeue.c
@@ -197,7 +197,7 @@ _print_job_steps( void )
(long) new_step_ptr->last_update);
if (params.format == NULL)
- params.format = "%10i %.9P %.8u %.9M %N";
+ params.format = "%10i %.8j %.9P %.8u %.9M %N";
if (params.format_list == NULL)
parse_format(params.format);
diff --git a/src/srun/allocate.c b/src/srun/allocate.c
index 22e4a9131a9..6579b34a4b8 100644
--- a/src/srun/allocate.c
+++ b/src/srun/allocate.c
@@ -465,6 +465,8 @@ _step_req_create(srun_job_t *j)
: (opt.nprocs*opt.cpus_per_task);
r->num_tasks = opt.nprocs;
r->node_list = j->nodelist;
+ r->network = opt.network;
+ r->name = opt.job_name;
r->relative = false; /* XXX fix this oneday */
switch (opt.distribution) {
--
GitLab