Minor improvements to logging

src/common/slurm_protocol_defs.h

@@ -322,7 +322,8 @@ typedef enum {
 	SRUN_EXEC,
 	SRUN_STEP_MISSING,
 	SRUN_REQUEST_SUSPEND,
-	SRUN_STEP_SIGNAL,	/* BluegeneQ: srun forwards signal to runjob */
+	SRUN_STEP_SIGNAL,	/* for launch plugins aprun, poe and runjob,
+				 * srun forwards signal to the launch command */
 
 	PMI_KVS_PUT_REQ = 7201,
 	PMI_KVS_PUT_RESP,

src/slurmd/slurmstepd/req.c

@@ -675,13 +675,11 @@ _handle_signal_task_local(int fd, slurmd_job_t *job, uid_t uid)
 	int signal;
 	int ltaskid; /* local task index */
 
-	debug("_handle_signal_task_local for job %u.%u",
-	      job->jobid, job->stepid);
-
 	safe_read(fd, &signal, sizeof(int));
 	safe_read(fd, &ltaskid, sizeof(int));
+	debug("_handle_signal_task_local for step=%u.%u uid=%d signal=%d",
+	      job->jobid, job->stepid, (int) uid, signal);
 
-	debug3("  uid = %d", uid);
 	if (uid != job->uid && !_slurm_authorized_user(uid)) {
 		debug("kill req from uid %ld for job %u.%u owned by uid %ld",
 		      (long)uid, job->jobid, job->stepid, (long)job->uid);
@@ -751,12 +749,9 @@ _handle_signal_container(int fd, slurmd_job_t *job, uid_t uid)
 	int sig;
 	static int msg_sent = 0;
 
-	debug("_handle_signal_container for job %u.%u",
-	      job->jobid, job->stepid);
-
 	safe_read(fd, &sig, sizeof(int));
-
-	debug3("  uid = %d", uid);
+	debug("_handle_signal_container for step=%u.%u uid=%d signal=%d",
+	      job->jobid, job->stepid, (int) uid, sig);
 	if (uid != job->uid && !_slurm_authorized_user(uid)) {
 		debug("kill container req from uid %ld for job %u.%u "
 		      "owned by uid %ld",
@@ -985,11 +980,10 @@ _handle_terminate(int fd, slurmd_job_t *job, uid_t uid)
 	int rc = SLURM_SUCCESS;
 	int errnum = 0;
 
-	debug("_handle_terminate for job %u.%u",
-	      job->jobid, job->stepid);
+	debug("_handle_terminate for step=%u.%u uid=%d",
+	      job->jobid, job->stepid, uid);
 	step_terminate_monitor_start(job->jobid, job->stepid);
 
-	debug3("  uid = %d", uid);
 	if (uid != job->uid && !_slurm_authorized_user(uid)) {
 		debug("terminate req from uid %ld for job %u.%u "
 		      "owned by uid %ld",
@@ -1159,8 +1153,8 @@ _handle_suspend(int fd, slurmd_job_t *job, uid_t uid)
 	int rc = SLURM_SUCCESS;
 	int errnum = 0;
 
-	debug("_handle_suspend for job %u.%u", job->jobid, job->stepid);
-	debug3("  uid = %d", uid);
+	debug("_handle_suspend for step=%u.%u uid=%d",
+	      job->jobid, job->stepi, (int) uid);
 	if (!_slurm_authorized_user(uid)) {
 		debug("job step suspend request from uid %ld for job %u.%u ",
 		      (long)uid, job->jobid, job->stepid);

testsuite/expect/test3.7

@@ -126,7 +126,8 @@ exec $bin_cc -o $file_prog ${file_prog}.c
 exec $bin_chmod 700 $file_prog
 
 #
-# Submit two jobs to the same node
+# Submit two jobs to the same node,.
+# The first job includes srun, second only the application
 #
 set srun_pid [spawn $sbatch -N1 -t2 --output=$file_out1 $file_prog_sh1]
 expect {
@@ -249,6 +250,7 @@ if {$exit_code == 0} {
 	exec $bin_rm -f $file_out1 $file_out2 $file_prog $file_prog_sh1 $file_prog_sh2
 	send_user "\nSUCCESS\n"
 } else {
-	send_user "\nFAILURE: May be due to use of gang scheduler, a race conditions, or the ProcTrack plugin not identifying the application as part of the job\n"
+	send_user "\nFAILURE: May be due to use of gang scheduler, a race conditions, or the ProcTrack plugin not identifying the application as part of the job.\n"
+	send_user "\nFAILURE: launch/poe and proctrack/pgid are incompatible.\n"
 }
 exit $exit_code