From 214fff55c5788162900cc38e14f58eacd8ae5ae0 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Tue, 20 Mar 2007 18:42:31 +0000
Subject: [PATCH] svn merge -r11200:11204
https://eris.llnl.gov/svn/slurm/branches/slurm-1.1 (Moab event notication
for partition state changes).
---
doc/html/schedplugins.shtml | 18 ++++++----
src/plugins/sched/backfill/backfill_wrapper.c | 8 +++++
src/plugins/sched/builtin/builtin_wrapper.c | 8 +++++
src/plugins/sched/gang/sched_gang.c | 8 +++++
src/plugins/sched/hold/hold_wrapper.c | 8 +++++
src/plugins/sched/wiki/sched_wiki.c | 8 +++++
src/plugins/sched/wiki2/event.c | 34 ++++++++++++-------
src/plugins/sched/wiki2/msg.c | 4 +--
src/plugins/sched/wiki2/msg.h | 4 +--
src/plugins/sched/wiki2/sched_wiki.c | 14 ++++++--
src/slurmctld/controller.c | 1 +
src/slurmctld/partition_mgr.c | 4 ++-
src/slurmctld/proc_req.c | 2 ++
src/slurmctld/sched_plugin.c | 15 ++++++++
src/slurmctld/sched_plugin.h | 5 +++
15 files changed, 114 insertions(+), 27 deletions(-)
diff --git a/doc/html/schedplugins.shtml b/doc/html/schedplugins.shtml
index 0b2a122ef46..915ab86a886 100644
--- a/doc/html/schedplugins.shtml
+++ b/doc/html/schedplugins.shtml
@@ -5,7 +5,7 @@
<h2> Overview</h2>
<p> This document describes SLURM scheduler plugins and the API that defines
them. It is intended as a resource to programmers wishing to write their own SLURM
-scheduler plugins. This is version 0 of the API.</p>
+scheduler plugins. This is version 1 of the API.</p>
<p>It is noteworthy that two different models are used for job scheduling.
The <b>backfill</b> scheduler lets SLURM establish the initial job priority
@@ -28,11 +28,11 @@ abbreviation for the type of scheduler. We recommend, for example:</p>
scheduling services. This is the default behavior and implements first-in-first-out scheduling.</li>
<li><b>backfill</b>—Raise the priority of jobs if doing so results in their starting earlier
without any delay in the expected initiation time of any higher priority job.</li>
-<li><b>wiki</b>—Use
+<li><b>wiki</b>—Uses
<a href="http://www.clusterresources.com/pages/products/maui-cluster-scheduler.php">
The Maui Scheduler</a> (Wiki version)
as an external entity to control SLURM job scheduling.</li>
-<li><b>wiki2</b>—Use
+<li><b>wiki2</b>—Uses
<a href="http://www.clusterresources.com/pages/products/moab-cluster-suite.php">
Moab Cluster Suite</a> as an external entity to control SLURM job scheduling.
Note that wiki2 is an expanded version of the wiki plugin with additional
@@ -103,7 +103,13 @@ by altering their priority and (optionally) list of required nodes.</p>
<p style="margin-left:.2in"><b>Returns</b>: The priority to be assigned to this job.</p>
<p class="commandline">void slurm_sched_plugin_job_is_pending (void);</p>
-<p style="margin-left:.2in"><b>Description</b>: Note that some job is pending execution..</p>
+<p style="margin-left:.2in"><b>Description</b>: Note that some job is pending execution.</p>
+<p style="margin-left:.2in"><b>Arguments</b>: None</p>
+<p style="margin-left:.2in"><b>Returns</b>: Nothing.</p>
+
+<p class="commandline">void slurm_sched_plugin_partition_change (void);</p>
+<p style="margin-left:.2in"><b>Description</b>: Note that some partition state change
+happened such as time or size limits.</p>
<p style="margin-left:.2in"><b>Arguments</b>: None</p>
<p style="margin-left:.2in"><b>Returns</b>: Nothing.</p>
<p class="footer"><a href="#top">top</a></p>
@@ -126,11 +132,11 @@ or NULL if no description found in this plugin.</p>
<p class="footer"><a href="#top">top</a></p>
<h2>Versioning</h2>
-<p> This document describes version 0 of the SLURM Scheduler API. Future
+<p> This document describes version 1 of the SLURM Scheduler API. Future
releases of SLURM may revise this API. A scheduler plugin conveys its ability
to implement a particular API version using the mechanism outlined for SLURM plugins.</p>
<p class="footer"><a href="#top">top</a></p>
-<p style="text-align:center;">Last modified 18 December 2006</p>
+<p style="text-align:center;">Last modified 20 March 2007</p>
<!--#include virtual="footer.txt"-->
diff --git a/src/plugins/sched/backfill/backfill_wrapper.c b/src/plugins/sched/backfill/backfill_wrapper.c
index ff60533507b..47a176c6673 100644
--- a/src/plugins/sched/backfill/backfill_wrapper.c
+++ b/src/plugins/sched/backfill/backfill_wrapper.c
@@ -157,6 +157,14 @@ void slurm_sched_plugin_job_is_pending( void )
run_backfill();
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ /* Empty. */
+}
+
/**************************************************************************/
/* TAG( slurm_sched_get_errno ) */
/**************************************************************************/
diff --git a/src/plugins/sched/builtin/builtin_wrapper.c b/src/plugins/sched/builtin/builtin_wrapper.c
index 95290aec238..422787b361b 100644
--- a/src/plugins/sched/builtin/builtin_wrapper.c
+++ b/src/plugins/sched/builtin/builtin_wrapper.c
@@ -103,6 +103,14 @@ void slurm_sched_plugin_job_is_pending( void )
/* Empty. */
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ /* Empty. */
+}
+
/**************************************************************************/
/* TAG( slurm_sched_get_errno ) */
/**************************************************************************/
diff --git a/src/plugins/sched/gang/sched_gang.c b/src/plugins/sched/gang/sched_gang.c
index ee0474387d8..483ed6e6e28 100644
--- a/src/plugins/sched/gang/sched_gang.c
+++ b/src/plugins/sched/gang/sched_gang.c
@@ -97,6 +97,14 @@ void slurm_sched_plugin_job_is_pending( void )
/* No action required */
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ /* No action required */
+}
+
/**************************************************************************/
/* TAG( slurm_sched_get_errno ) */
/**************************************************************************/
diff --git a/src/plugins/sched/hold/hold_wrapper.c b/src/plugins/sched/hold/hold_wrapper.c
index 5a4d952644e..bf1169162cc 100644
--- a/src/plugins/sched/hold/hold_wrapper.c
+++ b/src/plugins/sched/hold/hold_wrapper.c
@@ -113,6 +113,14 @@ void slurm_sched_plugin_job_is_pending( void )
/* Empty. */
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ /* Empty. */
+}
+
/**************************************************************************/
/* TAG( slurm_sched_get_errno ) */
/**************************************************************************/
diff --git a/src/plugins/sched/wiki/sched_wiki.c b/src/plugins/sched/wiki/sched_wiki.c
index 491775a93d4..1df597425d2 100644
--- a/src/plugins/sched/wiki/sched_wiki.c
+++ b/src/plugins/sched/wiki/sched_wiki.c
@@ -98,6 +98,14 @@ void slurm_sched_plugin_job_is_pending( void )
/* No action required */
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ /* Empty. */
+}
+
/**************************************************************************/
/* TAG( slurm_sched_plugin_reconfig ) */
/**************************************************************************/
diff --git a/src/plugins/sched/wiki2/event.c b/src/plugins/sched/wiki2/event.c
index 9999442dc80..bab1e8bba32 100644
--- a/src/plugins/sched/wiki2/event.c
+++ b/src/plugins/sched/wiki2/event.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* event.c - Moab event notification
*****************************************************************************
- * Copyright (C) 2006 The Regents of the University of California.
+ * Copyright (C) 2006-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -99,23 +99,35 @@ static void _close_fd(void)
/*
* event_notify - Notify Moab of some event
- * msg IN - event type, NULL to close connection
+ * event_code IN - message code to send Moab
+ * 1234 - job state change
+ * TBD - partition state change
+ * desc IN - event description
* RET 0 on success, -1 on failure
*/
-extern int event_notify(char *msg)
+extern int event_notify(int event_code, char *desc)
{
time_t now = time(NULL);
int rc = 0, retry = 2;
+ char *event_msg;
if (e_port == 0) {
/* Event notification disabled */
return 0;
}
- if (job_aggregation_time
- && (difftime(now, last_notify_time) < job_aggregation_time)) {
- debug("wiki event notification already sent recently");
- return 0;
+ if (event_code == 1234) {
+ if (job_aggregation_time
+ && (difftime(now, last_notify_time) < job_aggregation_time)) {
+ debug("wiki event notification already sent recently");
+ return 0;
+ }
+ event_msg = "1234\0";
+ } else if (event_code == 5) { /* actual value TBD */
+ event_msg = "5\0"; /* actual value TBD */
+ } else {
+ error("event_notify: invalid event code: %d", event_code);
+ return -1;
}
pthread_mutex_lock(&event_mutex);
@@ -128,12 +140,8 @@ extern int event_notify(char *msg)
break;
}
- /* Always send "1234\0" as the message
- * (we do not care if all of the message is sent,
- * just that some of it went through to wake up Moab)
- */
- if (write(event_fd, "1234", 5) > 0) {
- verbose("wiki event_notification sent: %s", msg);
+ if (write(event_fd, event_msg, (strlen(event_msg) + 1)) > 0) {
+ verbose("wiki event_notification sent: %s", desc);
last_notify_time = now;
rc = 0;
/* Dave Jackson says to leave the connection
diff --git a/src/plugins/sched/wiki2/msg.c b/src/plugins/sched/wiki2/msg.c
index 44fee14ea31..ca47133a787 100644
--- a/src/plugins/sched/wiki2/msg.c
+++ b/src/plugins/sched/wiki2/msg.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* msg.c - Message/communcation manager for Wiki plugin
*****************************************************************************
- * Copyright (C) 2006 The Regents of the University of California.
+ * Copyright (C) 2006-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -89,7 +89,7 @@ extern int spawn_msg_thread(void)
_msg_thread, NULL))
fatal("pthread_create %m");
- (void) event_notify("slurm startup");
+ (void) event_notify(1234, "Slurm startup");
slurm_attr_destroy(&thread_attr_msg);
thread_running = true;
pthread_mutex_unlock(&thread_flag_mutex);
diff --git a/src/plugins/sched/wiki2/msg.h b/src/plugins/sched/wiki2/msg.h
index 12fdec816b4..ac4e096b5f8 100644
--- a/src/plugins/sched/wiki2/msg.h
+++ b/src/plugins/sched/wiki2/msg.h
@@ -1,7 +1,7 @@
/*****************************************************************************\
* msg.h - Message/communcation manager for Wiki plugin
*****************************************************************************
- * Copyright (C) 2006 The Regents of the University of California.
+ * Copyright (C) 2006-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -94,7 +94,7 @@ extern uint16_t job_aggregation_time;
extern uint16_t kill_wait;
extern uint16_t use_host_exp;
-extern int event_notify(char *msg);
+extern int event_notify(int event_code, char *desc);
extern int spawn_msg_thread(void);
extern void term_msg_thread(void);
extern char * bitmap2wiki_node_name(bitstr_t *bitmap);
diff --git a/src/plugins/sched/wiki2/sched_wiki.c b/src/plugins/sched/wiki2/sched_wiki.c
index 231be3e996c..b7304c56873 100644
--- a/src/plugins/sched/wiki2/sched_wiki.c
+++ b/src/plugins/sched/wiki2/sched_wiki.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* sched_wiki.c - Wiki plugin for Moab and Maui schedulers.
*****************************************************************************
- * Copyright (C) 2006 The Regents of the University of California.
+ * Copyright (C) 2006-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -72,7 +72,7 @@ extern void fini( void )
/***************************************************************************/
extern int slurm_sched_plugin_schedule( void )
{
- (void) event_notify("Requested by slurm");
+ (void) event_notify(1234, "Requested by Slurm");
return SLURM_SUCCESS;
}
@@ -82,7 +82,7 @@ extern int slurm_sched_plugin_schedule( void )
/**************************************************************************/
extern uint32_t slurm_sched_plugin_initial_priority( uint32_t last_prio )
{
- (void) event_notify("Job submit");
+ (void) event_notify(1234, "Job submit");
if (init_prio_mode == PRIO_DECREMENT) {
if (last_prio >= 2)
return (last_prio - 1);
@@ -108,6 +108,14 @@ int slurm_sched_plugin_reconfig( void )
return parse_wiki_config();
}
+/**************************************************************************/
+/* TAG( slurm_sched_plugin_partition_change ) */
+/**************************************************************************/
+void slurm_sched_plugin_partition_change( void )
+{
+ (void) event_notify(1234, "Partition change");
+}
+
/**************************************************************************/
/* TAG( slurm_sched_get_errno ) */
/**************************************************************************/
diff --git a/src/slurmctld/controller.c b/src/slurmctld/controller.c
index 0028b840d9c..2cc71cd2543 100644
--- a/src/slurmctld/controller.c
+++ b/src/slurmctld/controller.c
@@ -521,6 +521,7 @@ static void *_slurmctld_signal_hand(void *no_data)
}
unlock_slurmctld(config_write_lock);
trigger_reconfig();
+ slurm_sched_partition_change();
break;
case SIGABRT: /* abort */
info("SIGABRT received");
diff --git a/src/slurmctld/partition_mgr.c b/src/slurmctld/partition_mgr.c
index 57210dc31b8..4023cf79717 100644
--- a/src/slurmctld/partition_mgr.c
+++ b/src/slurmctld/partition_mgr.c
@@ -4,7 +4,7 @@
* time stamp (last_part_update)
* $Id$
*****************************************************************************
- * Copyright (C) 2002 The Regents of the University of California.
+ * Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette@llnl.gov> et. al.
* UCRL-CODE-226842.
@@ -62,6 +62,7 @@
#include "src/slurmctld/locks.h"
#include "src/slurmctld/proc_req.h"
+#include "src/slurmctld/sched_plugin.h"
#include "src/slurmctld/slurmctld.h"
/* Global variables */
@@ -867,6 +868,7 @@ int update_part(update_part_msg_t * part_desc)
}
if (error_code == SLURM_SUCCESS) {
+ slurm_sched_partition_change(); /* notify sched plugin */
reset_job_priority(); /* free jobs */
if (select_g_block_init(part_list) != SLURM_SUCCESS )
error("failed to update node selection plugin state");
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index 65ae13b3660..0dd79befd2e 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -74,6 +74,7 @@
#include "src/slurmctld/locks.h"
#include "src/slurmctld/proc_req.h"
#include "src/slurmctld/read_config.h"
+#include "src/slurmctld/sched_plugin.h"
#include "src/slurmctld/slurmctld.h"
#include "src/slurmctld/state_save.h"
#include "src/slurmctld/trigger_mgr.h"
@@ -1429,6 +1430,7 @@ static void _slurm_rpc_reconfigure_controller(slurm_msg_t * msg)
info("_slurm_rpc_reconfigure_controller: completed %s",
TIME_STR);
slurm_send_rc_msg(msg, SLURM_SUCCESS);
+ slurm_sched_partition_change();
schedule(); /* has its own locks */
save_all_state();
}
diff --git a/src/slurmctld/sched_plugin.c b/src/slurmctld/sched_plugin.c
index cfc3c9d235c..44c00f15bf2 100644
--- a/src/slurmctld/sched_plugin.c
+++ b/src/slurmctld/sched_plugin.c
@@ -54,6 +54,7 @@ typedef struct slurm_sched_ops {
uint32_t (*initial_priority) ( uint32_t );
void (*job_is_pending) ( void );
int (*reconfig) ( void );
+ void (*partition_change) ( void );
int (*get_errno) ( void );
char * (*strerror) ( int );
} slurm_sched_ops_t;
@@ -88,6 +89,7 @@ slurm_sched_get_ops( slurm_sched_context_t *c )
"slurm_sched_plugin_initial_priority",
"slurm_sched_plugin_job_is_pending",
"slurm_sched_plugin_reconfig",
+ "slurm_sched_plugin_partition_change",
"slurm_sched_get_errno",
"slurm_sched_strerror"
};
@@ -268,6 +270,7 @@ slurm_sched_initial_priority( u_int32_t last_prio )
return (*(g_sched_context->ops.initial_priority))( last_prio );
}
+
/* *********************************************************************** */
/* TAG( slurm_sched_job_is_pending ) */
/* *********************************************************************** */
@@ -280,6 +283,18 @@ slurm_sched_job_is_pending( void )
(*(g_sched_context->ops.job_is_pending))();
}
+/* *********************************************************************** */
+/* TAG( slurm_sched_partition_change ) */
+/* *********************************************************************** */
+void
+slurm_sched_partition_change( void )
+{
+ if ( slurm_sched_init() < 0 )
+ return;
+
+ (*(g_sched_context->ops.partition_change))();
+}
+
/* *********************************************************************** */
/* TAG( slurm_sched_p_get_errno ) */
/* *********************************************************************** */
diff --git a/src/slurmctld/sched_plugin.h b/src/slurmctld/sched_plugin.h
index 6d24ac006bb..4dbb99ff9db 100644
--- a/src/slurmctld/sched_plugin.h
+++ b/src/slurmctld/sched_plugin.h
@@ -80,6 +80,11 @@ uint32_t slurm_sched_initial_priority( uint32_t max_prio );
*/
void slurm_sched_job_is_pending( void );
+/*
+ * Note that some partition state change happened.
+ */
+void slurm_sched_partition_change( void );
+
/*
* Return any plugin-specific error number
*/
--
GitLab