Allow for almost any node state flags to be in SuspendExcStates

Continuation of 94ce256.

Bug 15184

doc/html/elastic_computing.shtml

@@ -141,8 +141,10 @@ Multiple partitions may be specified using a comma separator.
 By default, no nodes are excluded.
 <dt><b>SuspendExcStates</b>
 <dd>Specifies node states that are not to be powered down automatically.
-Valid options include Down, Drain and Planned. By default, any of these states,
-if idle for <b>SuspendTime</b>, would be powered down.
+Valid states include CLOUD, DOWN, DRAIN, DYNAMIC_FUTURE, DYNAMIC_NORM, FAIL,
+INVALID_REG, MAINTENANCE, NOT_RESPONDING, PERFCTRS, PLANNED, and RESERVED.
+By default, any of these states, if idle for <b>SuspendTime</b>, would be
+powered down.
 <dt><b>SuspendProgram</b>
 <dd>The program executed when a node is no longer required and can be
 relinquished to the cloud.

doc/html/power_save.shtml

@@ -116,9 +116,11 @@ Multiple partitions may be specified using a comma separator.
 By default, no nodes are excluded.</li>
 
 <li><b>SuspendExcStates</b>:
-Specifies node states that are not to be powered down automatically. Valid
-options include Down, Drain and Planned. By default, any of these states, if
-idle for <b>SuspendTime</b>, would be powered down.
+Specifies node states that are not to be powered down automatically.
+Valid states include CLOUD, DOWN, DRAIN, DYNAMIC_FUTURE, DYNAMIC_NORM, FAIL,
+INVALID_REG, MAINTENANCE, NOT_RESPONDING, PERFCTRS, PLANNED, and RESERVED.
+By default, any of these states, if idle for <b>SuspendTime</b>, would be
+powered down.
 
 <li><b>BatchStartTimeout</b>:
 Specifies how long to wait after a batch job start request is issued

doc/man/man5/slurm.conf.5

@@ -5271,7 +5271,8 @@ By default no nodes are excluded.
 .TP
 \fBSuspendExcStates\fR
 Specifies node states that are not to be powered down automatically.
-Valid options include Down, Drain and Planned.
+Valid states include CLOUD, DOWN, DRAIN, DYNAMIC_FUTURE, DYNAMIC_NORM, FAIL,
+INVALID_REG, MAINTENANCE, NOT_RESPONDING, PERFCTRS, PLANNED, and RESERVED.
 By default, any of these states, if idle for SuspendTime, would be powered down.
 .IP
 

src/common/slurm_protocol_defs.c

@@ -3789,6 +3789,21 @@ extern char *node_state_string_complete(uint32_t state)
 	return state_str;
 }
 
+extern uint32_t parse_node_state_flag(char *flag_str)
+{
+	int flag_str_len = strlen(flag_str);
+
+	for (int i = 0; i < ARRAY_SIZE(node_state_flags); i++) {
+		if (!xstrncasecmp(flag_str, node_state_flags[i].str,
+				  MIN(flag_str_len,
+				      strlen(node_state_flags[i].str)))) {
+			return node_state_flags[i].flag;
+		}
+	}
+
+	return 0; /* No flags */
+}
+
 extern char *node_state_string(uint32_t inx)
 {
 	int  base            = (inx & NODE_STATE_BASE);

src/common/slurm_protocol_defs.h

@@ -1704,6 +1704,11 @@ extern char *node_state_string_compact(uint32_t inx);
  */
 extern char *node_state_string_complete(uint32_t inx);
 
+/*
+ * Return first node state flag that matches the string
+ */
+extern uint32_t parse_node_state_flag(char *flag_str);
+
 extern uint16_t power_flags_id(const char *power_flags);
 extern char    *power_flags_str(uint16_t power_flags);
 

src/slurmctld/power_save.c

@@ -110,7 +110,8 @@ List partial_node_list;
 bitstr_t *exc_node_bitmap = NULL;
 
 /* Possible SuspendExcStates */
-static bool suspend_exc_down, suspend_exc_drain, suspend_exc_planned;
+static bool suspend_exc_down;
+static uint32_t suspend_exc_state_flags;
 
 int   suspend_cnt,   resume_cnt;
 float suspend_cnt_f, resume_cnt_f;
@@ -188,20 +189,38 @@ static int _list_part_node_lists(void *x, void *arg)
 static void _parse_exc_states(void)
 {
 	char *buf, *tok, *saveptr;
-	int tok_len;
+	/* Flags in _node_state_suspendable() are already excluded */
+	uint32_t excludable_state_flags = NODE_STATE_CLOUD |
+					  NODE_STATE_DRAIN |
+					  NODE_STATE_DYNAMIC_FUTURE |
+					  NODE_STATE_DYNAMIC_NORM |
+					  NODE_STATE_FAIL |
+					  NODE_STATE_INVALID_REG |
+					  NODE_STATE_MAINT |
+					  NODE_STATE_NET |
+					  NODE_STATE_NO_RESPOND |
+					  NODE_STATE_PLANNED |
+					  NODE_STATE_RES;
 
 	buf = xstrdup(exc_states);
 	for (tok = strtok_r(buf, ",", &saveptr); tok;
 	     tok = strtok_r(NULL, ",", &saveptr)) {
-		tok_len = strlen(tok);
-		if (!xstrncasecmp(tok, "Down", MAX(tok_len, 2)))
+		uint32_t flag = 0;
+
+		/* Base node states */
+		if (!xstrncasecmp(tok, "DOWN", MAX(strlen(tok), 2))){
 			suspend_exc_down = true;
-		else if (!xstrncasecmp(tok, "Drained", MAX(tok_len, 2)))
-			suspend_exc_drain = true;
-		else if (!xstrncasecmp(tok, "Planned", MAX(tok_len, 1)))
-			suspend_exc_planned = true;
-		else
-			error("Invalid SuspendExcState %s", tok);
+			continue;
+		}
+
+		/* Flag node states */
+		flag = parse_node_state_flag(tok);
+		if (flag & excludable_state_flags) {
+			suspend_exc_state_flags |= flag;
+			continue;
+		}
+
+		error("Invalid SuspendExcState %s", tok);
 	}
 	xfree(buf);
 }
@@ -235,9 +254,7 @@ static bool _node_state_should_suspend(node_record_t *node_ptr)
 	/* SuspendExcStates */
 	if (suspend_exc_down && IS_NODE_DOWN(node_ptr))
 		return false;
-	if (suspend_exc_drain && IS_NODE_DRAIN(node_ptr))
-		return false;
-	if (suspend_exc_planned && (node_ptr->node_state & NODE_STATE_PLANNED))
+	if (suspend_exc_state_flags & node_ptr->node_state)
 		return false;
 
 	return true;
@@ -701,8 +718,7 @@ static void _clear_power_config(void)
 	xfree(exc_parts);
 	xfree(exc_states);
 	suspend_exc_down = false;
-	suspend_exc_drain = false;
-	suspend_exc_planned = false;
+	suspend_exc_state_flags = 0;
 	FREE_NULL_BITMAP(exc_node_bitmap);
 	FREE_NULL_LIST(partial_node_list);
 }