diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index d36f2061ec495820ce5cee4a8cfec827c91dd4f0..93016eae3cf6fae211680657db55c1339aa6b8a5 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -145,7 +145,9 @@ inline static void _slurm_rpc_dump_sicp(slurm_msg_t * msg);
inline static void _slurm_rpc_dump_spank(slurm_msg_t * msg);
inline static void _slurm_rpc_dump_stats(slurm_msg_t * msg);
inline static void _slurm_rpc_end_time(slurm_msg_t * msg);
-inline static void _slurm_rpc_epilog_complete(slurm_msg_t * msg);
+inline static void _slurm_rpc_epilog_complete(slurm_msg_t * msg,
+ bool *run_scheduler,
+ bool running_composite);
inline static void _slurm_rpc_get_shares(slurm_msg_t *msg);
inline static void _slurm_rpc_get_topo(slurm_msg_t * msg);
inline static void _slurm_rpc_get_priority_factors(slurm_msg_t *msg);
@@ -311,7 +313,8 @@ void slurmctld_req(slurm_msg_t *msg, connection_arg_t *arg)
slurm_free_part_info_request_msg(msg->data);
break;
case MESSAGE_EPILOG_COMPLETE:
- _slurm_rpc_epilog_complete(msg);
+ i = 0;
+ _slurm_rpc_epilog_complete(msg, (bool *)&i, 0);
slurm_free_epilog_complete_msg(msg->data);
break;
case REQUEST_CANCEL_JOB_STEP:
@@ -1669,7 +1672,9 @@ static void _slurm_rpc_dump_partitions(slurm_msg_t * msg)
/* _slurm_rpc_epilog_complete - process RPC noting the completion of
* the epilog denoting the completion of a job it its entirety */
-static void _slurm_rpc_epilog_complete(slurm_msg_t * msg)
+static void _slurm_rpc_epilog_complete(slurm_msg_t *msg,
+ bool *run_scheduler,
+ bool running_composite)
{
static int active_rpc_cnt = 0;
static time_t config_update = 0;
@@ -1681,7 +1686,6 @@ static void _slurm_rpc_epilog_complete(slurm_msg_t * msg)
uid_t uid = g_slurm_auth_get_uid(msg->auth_cred, NULL);
epilog_complete_msg_t *epilog_msg =
(epilog_complete_msg_t *) msg->data;
- bool run_scheduler = false;
struct job_record *job_ptr;
char jbuf[JBUFSIZ];
@@ -1694,26 +1698,35 @@ static void _slurm_rpc_epilog_complete(slurm_msg_t * msg)
}
job_ptr = find_job_record(epilog_msg->job_id);
- if (config_update != slurmctld_conf.last_update) {
- char *sched_params = slurm_get_sched_params();
- defer_sched = (sched_params && strstr(sched_params,"defer"));
- xfree(sched_params);
- config_update = slurmctld_conf.last_update;
- }
+ /* Only throttle on none composite messages, the lock should
+ * already be set earlier. */
+ if (!running_composite) {
+ if (config_update != slurmctld_conf.last_update) {
+ char *sched_params = slurm_get_sched_params();
+ defer_sched = (sched_params &&
+ strstr(sched_params, "defer"));
+ xfree(sched_params);
+ config_update = slurmctld_conf.last_update;
+ }
- _throttle_start(&active_rpc_cnt);
+ _throttle_start(&active_rpc_cnt);
+ lock_slurmctld(job_write_lock);
+ }
if (slurmctld_conf.debug_flags & DEBUG_FLAG_ROUTE)
info("_slurm_rpc_epilog_complete: "
"node_name = %s, job_id = %u", epilog_msg->node_name,
epilog_msg->job_id);
- lock_slurmctld(job_write_lock);
if (job_epilog_complete(epilog_msg->job_id, epilog_msg->node_name,
epilog_msg->return_code))
- run_scheduler = true;
- unlock_slurmctld(job_write_lock);
- _throttle_fini(&active_rpc_cnt);
+ *run_scheduler = true;
+
+ if (!running_composite) {
+ unlock_slurmctld(job_write_lock);
+ _throttle_fini(&active_rpc_cnt);
+ }
+
END_TIMER2("_slurm_rpc_epilog_complete");
if (epilog_msg->return_code)
@@ -1727,7 +1740,7 @@ static void _slurm_rpc_epilog_complete(slurm_msg_t * msg)
epilog_msg->node_name, TIME_STR);
/* Functions below provide their own locking */
- if (run_scheduler) {
+ if (!running_composite && *run_scheduler) {
/*
* In defer mode, avoid triggering the scheduler logic
* for every epilog complete message.
@@ -5333,20 +5346,16 @@ static void _slurm_rpc_comp_msg_list(composite_msg_t * comp_msg,
bool *run_scheduler,
List msg_list_in)
{
- uid_t uid;
- epilog_complete_msg_t *epilog_msg;
- composite_msg_t *comp_resp_msg;
- struct job_record *job_ptr;
- char jbuf[JBUFSIZ];
- DEF_TIMERS;
-
ListIterator itr;
slurm_msg_t *next_msg;
composite_msg_t *ncomp_msg;
+ composite_msg_t *comp_resp_msg;
itr = list_iterator_create(comp_msg->msg_list);
while ((next_msg = list_next(itr))) {
- uid = g_slurm_auth_get_uid(next_msg->auth_cred, NULL);
+ /* The ret_list is used by slurm_send_rc_msg to house
+ replys going back to the nodes.
+ */
FREE_NULL_LIST(next_msg->ret_list);
next_msg->ret_list = msg_list_in;
switch (next_msg->msg_type) {
@@ -5389,40 +5398,7 @@ static void _slurm_rpc_comp_msg_list(composite_msg_t * comp_msg,
slurm_free_composite_msg(comp_resp_msg);
break;
case MESSAGE_EPILOG_COMPLETE:
- epilog_msg = (epilog_complete_msg_t *) next_msg->data;
- START_TIMER;
- if (slurmctld_conf.debug_flags & DEBUG_FLAG_ROUTE)
- info("Processing embedded "
- "MESSAGE_EPILOG_COMPLETE uid=%d %d",
- uid, next_msg->msg_index);
- if (!validate_slurm_user(uid)) {
- error("Security violation, EPILOG_COMPLETE RPC "
- "from uid=%d", uid);
- return;
- }
-
- job_ptr = find_job_record(epilog_msg->job_id);
-
- if (job_epilog_complete(epilog_msg->job_id,
- epilog_msg->node_name,
- epilog_msg->return_code))
- *run_scheduler = true;
- END_TIMER2("_slurm_rpc_comp_msg_list");
-
- if (epilog_msg->return_code)
- error("%s: epilog error %s Node=%s Err=%s %s",
- __func__, jobid2str(job_ptr, jbuf,
- sizeof(jbuf)),
- epilog_msg->node_name,
- slurm_strerror(epilog_msg->return_code),
- TIME_STR);
- else
- if (slurmctld_conf.debug_flags &
- DEBUG_FLAG_ROUTE)
- info("%s: %s Node=%s %s",
- __func__, jobid2str(job_ptr, jbuf,
- sizeof(jbuf)),
- epilog_msg->node_name, TIME_STR);
+ _slurm_rpc_epilog_complete(next_msg, run_scheduler, 1);
slurm_send_rc_msg(next_msg, SLURM_SUCCESS);
break;
case MESSAGE_NODE_REGISTRATION_STATUS: