extern Buf slurm_persist_make_rc_msg(slurm_persist_conn_t *persist_conn,
uint32_t rc, char *comment,
uint16_t ret_info)
{
persist_rc_msg_t msg;
persist_msg_t resp;
memset(&msg, 0, sizeof(persist_rc_msg_t));
memset(&resp, 0, sizeof(persist_msg_t));
msg.rc = rc;
msg.comment = comment;
msg.ret_info = ret_info;
resp.msg_type = PERSIST_RC;
resp.data = &msg;
return slurm_persist_msg_pack(persist_conn, &resp);
}
/* Send an RPC to the SlurmDBD. Do not wait for the reply. The RPC
* will be queued and processed later if the SlurmDBD is not responding.
* NOTE: slurm_open_slurmdbd_conn() must have been called with callbacks set
*
* Returns SLURM_SUCCESS or an error code */
extern int send_slurmdbd_msg(uint16_t rpc_version, slurmdbd_msg_t *req)
{
Buf buffer;
int cnt, rc = SLURM_SUCCESS;
static time_t syslog_time = 0;
static int max_agent_queue = 0;
/*
* Whatever our max job count is multiplied by 2 plus node count
* multiplied by 4 or MAX_AGENT_QUEUE which ever is bigger.
*/
if (!max_agent_queue)
max_agent_queue =
MAX(MAX_AGENT_QUEUE,
((slurmctld_conf.max_job_cnt * 2) +
(node_record_count * 4)));
buffer = slurm_persist_msg_pack(
slurmdbd_conn, (persist_msg_t *)req);
if (!buffer) /* pack error */
return SLURM_ERROR;
slurm_mutex_lock(&agent_lock);
if ((agent_tid == 0) || (agent_list == NULL)) {
_create_agent();
if ((agent_tid == 0) || (agent_list == NULL)) {
slurm_mutex_unlock(&agent_lock);
free_buf(buffer);
return SLURM_ERROR;
}
}
cnt = list_count(agent_list);
if ((cnt >= (max_agent_queue / 2)) &&
(difftime(time(NULL), syslog_time) > 120)) {
/* Record critical error every 120 seconds */
syslog_time = time(NULL);
error("slurmdbd: agent queue filling (%d), RESTART SLURMDBD NOW",
cnt);
syslog(LOG_CRIT, "*** RESTART SLURMDBD NOW ***");
if (slurmdbd_conn->trigger_callbacks.dbd_fail)
(slurmdbd_conn->trigger_callbacks.dbd_fail)();
}
if (cnt == (max_agent_queue - 1))
cnt -= _purge_step_req();
if (cnt == (max_agent_queue - 1))
cnt -= _purge_job_start_req();
if (cnt < max_agent_queue) {
if (list_enqueue(agent_list, buffer) == NULL)
fatal("list_enqueue: memory allocation failure");
} else {
error("slurmdbd: agent queue is full (%u), discarding %s:%u request",
cnt,
slurmdbd_msg_type_2_str(req->msg_type, 1),
req->msg_type);
if (slurmdbd_conn->trigger_callbacks.acct_full)
(slurmdbd_conn->trigger_callbacks.acct_full)();
free_buf(buffer);
rc = SLURM_ERROR;
}
slurm_cond_broadcast(&agent_cond);
slurm_mutex_unlock(&agent_lock);
return rc;
}
