diff --git a/NEWS b/NEWS
index 598122b278818f2cdd5e771e322bbc01c5d3b9fe..8ceb24d8d97ce6fe5c37b8c45e6fb2aebf05cf9a 100644
--- a/NEWS
+++ b/NEWS
@@ -82,6 +82,12 @@ documents those changes that are of interest to users and admins.
the code)
-- Added support for OSX build.
+* Changes in SLURM 1.1.18
+=========================
+ - In sched/wiki2, add support for EHost and EHostBackup configuration
+ parameters in wiki.conf file
+ - In sched/wiki2, fix memory management bug for JOBWILLRUN command.
+
* Changes in SLURM 1.1.17
=========================
- BLUEGENE - fix to make dynamic partitioning not go create block where
diff --git a/doc/man/man5/wiki.conf.5 b/doc/man/man5/wiki.conf.5
index 56d49ff55103e7510f00efb74ba1b122fdbc7ad7..772ff3c0f46dbadbd9a4d2c36310875243c68df7 100644
--- a/doc/man/man5/wiki.conf.5
+++ b/doc/man/man5/wiki.conf.5
@@ -1,4 +1,4 @@
-.TH "wiki.conf" "5" "September 2006" "wiki.conf 1.1" "Slurm configuration file"
+.TH "wiki.conf" "5" "October 2006" "wiki.conf 1.1" "Slurm configuration file"
.SH "NAME"
wiki.conf \- Slurm configuration file for wiki scheduler plugin
.SH "DESCRIPTION"
@@ -21,6 +21,20 @@ Authentication key for communications.
This numeric value should match KEY configured in the
\fBmoab\-private.cnf\fR file.
+.TP
+\fBEHost\fR
+Name the computer on which Moab server executes.
+It is used in establishing a communications path for event notification.
+By default the \fBEHost\fR will be identical in value to the
+\fBControlAddr\fR configured in slurm.conf.
+
+.TP
+\fBEHostBackup\fR
+Name the computer on which the backup Moab server executes.
+It is used in establishing a communications path for event notification.
+There is no default value for \fBEHostBackup\fR (no backup
+controller is configured).
+
.TP
\fBEPort\fR
Port to be used to notify Moab of events (job submitted to Slurm,
@@ -77,6 +91,12 @@ AuthKey=1234
.br
EPort=15017
.br
+# Moab event notification hosts, where Moab executes
+.br
+EHost=tux0
+.br
+EHostBackup=tux1
+.br
# Moab event notifcation throttle, matches JOBAGGREGATIONTIME
.br
# in moab.cfg (integer value in seconds)
diff --git a/src/plugins/sched/wiki2/event.c b/src/plugins/sched/wiki2/event.c
index fe7817d12f1c69cb27c68ca3ac4eb421c1ce94b1..4e57f732b005442e1a6075a3f5ac02d24809c2a4 100644
--- a/src/plugins/sched/wiki2/event.c
+++ b/src/plugins/sched/wiki2/event.c
@@ -40,9 +40,8 @@
static pthread_mutex_t event_mutex = PTHREAD_MUTEX_INITIALIZER;
static time_t last_notify_time = (time_t) 0;
-static slurm_addr moab_event_addr;
+static slurm_addr moab_event_addr, moab_event_addr_bu;
static int event_addr_set = 0;
-static char * control_addr = NULL;
/*
* event_notify - Notify Moab of some event
@@ -69,16 +68,31 @@ extern int event_notify(char *msg)
pthread_mutex_lock(&event_mutex);
if (event_addr_set == 0) {
/* Identify address for socket connection */
- slurm_ctl_conf_t *conf = slurm_conf_lock();
- control_addr = xstrdup(conf->control_addr);
- slurm_conf_unlock();
- slurm_set_addr(&moab_event_addr, e_port, control_addr);
+ if (e_host[0] == '\0') {
+ slurm_ctl_conf_t *conf = slurm_conf_lock();
+ strncpy(e_host, conf->control_addr,
+ sizeof(e_host));
+ slurm_conf_unlock();
+ }
+ slurm_set_addr(&moab_event_addr, e_port, e_host);
event_addr_set = 1;
+ if (e_host_bu[0] != '\0') {
+ slurm_set_addr(&moab_event_addr_bu, e_port,
+ e_host_bu);
+ event_addr_set = 2;
+ }
}
event_fd = slurm_open_msg_conn(&moab_event_addr);
+ if ((event_fd == -1) && (event_addr_set == 2))
+ event_fd = slurm_open_msg_conn(&moab_event_addr_bu);
if (event_fd == -1) {
+ char *host_name;
+ if (event_addr_set == 2)
+ host_name = e_host_bu;
+ else
+ host_name = e_host;
error("Unable to open wiki event port %s:%u: %m",
- control_addr, e_port);
+ host_name, e_port);
pthread_mutex_unlock(&event_mutex);
return -1;
}
diff --git a/src/plugins/sched/wiki2/job_will_run.c b/src/plugins/sched/wiki2/job_will_run.c
index ce260c6331a7312debbf9fb65230cede74bc97a6..05293819425106e4127589a6e2a8a33bc045a30e 100644
--- a/src/plugins/sched/wiki2/job_will_run.c
+++ b/src/plugins/sched/wiki2/job_will_run.c
@@ -108,9 +108,11 @@ static int _will_run_test(uint32_t jobid, char *hostlist,
char *new_node_list, *picked_node_list = NULL;
bitstr_t *new_bitmap, *save_exc_bitmap, *save_req_bitmap;
uint32_t save_prio;
- static char *reply_msg;
- static int reply_msg_size = 0;
bitstr_t *picked_node_bitmap = NULL;
+ /* Just create a big static message buffer to avoid dealing with
+ * xmalloc/xfree. We'll switch to compressed node naming soon
+ * and this buffer can be set smaller then. */
+ static char reply_msg[16384];
lock_slurmctld(job_write_lock);
job_ptr = find_job_record(jobid);
@@ -172,21 +174,19 @@ static int _will_run_test(uint32_t jobid, char *hostlist,
if (picked_node_bitmap) {
picked_node_list = bitmap2wiki_node_name(picked_node_bitmap);
i = strlen(picked_node_list);
- if ((i + 64) > reply_msg_size) {
- reply_msg_size = i + 1024;
- xrealloc(reply_msg, reply_msg_size);
- }
+ if ((i + 64) > sizeof(reply_msg))
+ error("wiki: will_run buffer overflow");
}
if (rc == SLURM_SUCCESS) {
*err_code = 0;
- snprintf(reply_msg, reply_msg_size,
+ snprintf(reply_msg, sizeof(reply_msg),
"SC=0 Job %d runnable now TASKLIST:%s",
jobid, picked_node_list);
*err_msg = reply_msg;
} else if (rc == ESLURM_NODES_BUSY) {
*err_code = 1;
- snprintf(reply_msg, reply_msg_size,
+ snprintf(reply_msg, sizeof(reply_msg),
"SC=1 Job %d runnable later TASKLIST:%s",
jobid, picked_node_list);
*err_msg = reply_msg;
@@ -195,7 +195,7 @@ static int _will_run_test(uint32_t jobid, char *hostlist,
error("wiki: job %d never runnable on hosts=%s %s",
jobid, new_node_list, err_str);
*err_code = -740;
- snprintf(reply_msg, reply_msg_size,
+ snprintf(reply_msg, sizeof(reply_msg),
"SC=-740 Job %d not runable: %s",
jobid, err_str);
*err_msg = reply_msg;
diff --git a/src/plugins/sched/wiki2/msg.c b/src/plugins/sched/wiki2/msg.c
index 40916cffa5351fa3799cbc4a0a6099a093022811..e933da77e80bae2eb0371446724195ec439cf303 100644
--- a/src/plugins/sched/wiki2/msg.c
+++ b/src/plugins/sched/wiki2/msg.c
@@ -48,6 +48,8 @@ static int err_code;
/* Global configuration parameters */
char auth_key[KEY_SIZE] = "";
+char e_host[E_HOST_SIZE] = "";
+char e_host_bu[E_HOST_SIZE] = "";
uint16_t e_port = 0;
uint16_t job_aggregation_time = 10; /* Default value is 10 seconds */
int init_prio_mode = PRIO_HOLD;
@@ -194,6 +196,8 @@ static void _parse_wiki_config(void)
{
s_p_options_t options[] = {
{"AuthKey", S_P_STRING},
+ {"EHost", S_P_STRING},
+ {"EHostBackup", S_P_STRING},
{"EPort", S_P_UINT16},
{"JobAggregationTime", S_P_UINT16},
{"JobPriority", S_P_STRING},
@@ -220,6 +224,14 @@ static void _parse_wiki_config(void)
strncpy(auth_key, key, sizeof(auth_key));
xfree(key);
}
+ if ( s_p_get_string(&key, "EHost", tbl)) {
+ strncpy(e_host, key, sizeof(e_host));
+ xfree(key);
+ }
+ if ( s_p_get_string(&key, "EHostBackup", tbl)) {
+ strncpy(e_host_bu, key, sizeof(e_host_bu));
+ xfree(key);
+ }
s_p_get_uint16(&e_port, "EPort", tbl);
s_p_get_uint16(&job_aggregation_time, "JobAggregationTime", tbl);
@@ -459,8 +471,7 @@ static void _proc_msg(slurm_fd new_fd, char *msg)
} else if (strncmp(cmd_ptr, "JOBRELEASETASK", 14) == 0) {
job_release_task(cmd_ptr, &err_code, &err_msg);
} else if (strncmp(cmd_ptr, "JOBWILLRUN", 10) == 0) {
- if (!job_will_run(cmd_ptr, &err_code, &err_msg))
- goto free_resp_msg;
+ job_will_run(cmd_ptr, &err_code, &err_msg);
} else if (strncmp(cmd_ptr, "JOBMODIFY", 9) == 0) {
job_modify_wiki(cmd_ptr, &err_code, &err_msg);
} else if (strncmp(cmd_ptr, "JOBSIGNAL", 9) == 0) {
diff --git a/src/plugins/sched/wiki2/msg.h b/src/plugins/sched/wiki2/msg.h
index 431918187953d239e2a817ce12c803b196e06b77..c46254385d02e93896441ec83e83de4d4eaac4e9 100644
--- a/src/plugins/sched/wiki2/msg.h
+++ b/src/plugins/sched/wiki2/msg.h
@@ -81,11 +81,14 @@
#include "src/common/xstring.h"
/* Global configuration parameters */
+#define E_HOST_SIZE 256
#define KEY_SIZE 32
#define PRIO_HOLD 0
#define PRIO_DECREMENT 1
extern int init_prio_mode;
extern char auth_key[KEY_SIZE];
+extern char e_host[E_HOST_SIZE];
+extern char e_host_bu[E_HOST_SIZE];
extern uint16_t e_port;
extern uint16_t job_aggregation_time;
extern uint16_t use_host_exp;