lei.chen/myxce
1
0
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix scheduler notification data pointer ownership and better trace log for unsatisfy_()

Browse source
This commit is contained in:
Rongsong Shen 2026-05-12 13:45:58 +00:00
parent a07a74a943
commit 9f6741c2c4
2 changed files with 21 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

8
src/scheduler/src/handler.c
View file

@ -294,18 +294,26 @@ int sched_handle_notify(xce_scheduler_t* sched, xce_notify_data_t* notify)
case NotifyData_data_schedUpdateJob: case NotifyData_data_schedUpdateJob:
rc = sched_handle_sched_update_job(sched, rc = sched_handle_sched_update_job(sched,
notify->data.sched_update_job); notify->data.sched_update_job);
if (rc >= 0)
notify->data.sched_update_job = NULL;
break; break;
case NotifyData_data_schedUpdateQueue: case NotifyData_data_schedUpdateQueue:
rc = sched_handle_sched_update_queue(sched, rc = sched_handle_sched_update_queue(sched,
notify->data.sched_update_queue); notify->data.sched_update_queue);
if (rc >= 0)
notify->data.sched_update_queue = NULL;
break; break;
case NotifyData_data_schedUpdateResvals: case NotifyData_data_schedUpdateResvals:
rc = sched_handle_sched_update_resvals( rc = sched_handle_sched_update_resvals(
sched, notify->data.sched_update_resvals); sched, notify->data.sched_update_resvals);
if (rc >= 0)
notify->data.sched_update_resvals = NULL;
break; break;
case NotifyData_data_schedJobRusages: case NotifyData_data_schedJobRusages:
rc = sched_handle_sched_job_rusage(sched, rc = sched_handle_sched_job_rusage(sched,
notify->data.sched_job_rusages); notify->data.sched_job_rusages);
if (rc >= 0)
notify->data.sched_job_rusages = NULL;
break; break;
default: default:
rc = -EINVAL; rc = -EINVAL;

19
src/scheduler/src/resource.c
View file

@ -93,9 +93,9 @@ int put_resvals(zarray_t* array, int nres, int* idxs, float* vals)
return nres; return nres;
} }
static int unsatisfy_(int nres, int* idxs, float* total, float* inuse, static int unsatisfy_(xce_scheduler_t* sched, int nres, int* idxs, float* total,
float* uncomitted, float* wanted, int* unmeet_idxs, float* inuse, float* uncomitted, float* wanted,
float* unmeet_vals) int* unmeet_idxs, float* unmeet_vals)
{ {
int i, j; int i, j;
int count; int count;
@ -116,8 +116,15 @@ static int unsatisfy_(int nres, int* idxs, float* total, float* inuse,
j = 0; j = 0;
for (i = 0; i < nres; i++) { for (i = 0; i < nres; i++) {
if (total[i] < inuse[i] + uncomitted[i] + wanted[i]) { if (total[i] < inuse[i] + uncomitted[i] + wanted[i]) {
xce_resource_t* r = NULL;
float need = inuse[i] + uncomitted[i] + wanted[i] - total[i];
vget(sched->resmgr, idxs[i], (void**)&r);
TRACE("%s unmet resource %d (%s) shortfall %f", __func__, idxs[i],
r ? r->name : "?", need);
unmeet_idxs[j] = idxs[i]; unmeet_idxs[j] = idxs[i];
unmeet_vals[j] = inuse[i] + wanted[i] + uncomitted[i] - total[i]; unmeet_vals[j] = need;
j++; j++;
} }
} }
@ -249,7 +256,7 @@ static int sched_check_shared_satisfy_(xce_scheduler_t* sched, xvec_t* total,
return rc; return rc;
} }
rc = unsatisfy_(nres, (int*)wanted->idxs.data, (float*)total->data, rc = unsatisfy_(sched, nres, (int*)wanted->idxs.data, (float*)total->data,
(float*)inuse->data, (float*)uncomitted->data, (float*)inuse->data, (float*)uncomitted->data,
(float*)wanted->vals.data, (int*)unmeet->idxs.data, (float*)wanted->vals.data, (int*)unmeet->idxs.data,
(float*)unmeet->vals.data); (float*)unmeet->vals.data);
@ -324,7 +331,7 @@ static int sched_check_host_satisfy_(xce_scheduler_t* sched,
return rc; return rc;
} }
rc = unsatisfy_(nres, (int*)wanted->idxs.data, (float*)total->data, rc = unsatisfy_(sched, nres, (int*)wanted->idxs.data, (float*)total->data,
(float*)inuse->data, (float*)uncomitted->data, (float*)inuse->data, (float*)uncomitted->data,
(float*)wanted->vals.data, (int*)unmeet->idxs.data, (float*)wanted->vals.data, (int*)unmeet->idxs.data,
(float*)unmeet->vals.data); (float*)unmeet->vals.data);