|
|
@ -349,18 +349,21 @@ def get_table_res(data_list): |
|
|
|
|
|
|
|
return ''.join(result_str) |
|
|
|
|
|
|
|
def keep_check_one(server: dict, shared_data_list: list, server_idx: int, interval: float): |
|
|
|
def keep_check_one(server: dict, shared_data_list: list, server_idx: int, interval: float, re_connect_time: float=5): |
|
|
|
# 循环连接 |
|
|
|
while run_realtime: |
|
|
|
try: |
|
|
|
# 建立SSH连接 |
|
|
|
client = paramiko.SSHClient() |
|
|
|
client.set_missing_host_key_policy(paramiko.AutoAddPolicy()) |
|
|
|
client.connect(server['ip'], port=server['port'], username=server['username'], password=server.get('password', None), key_filename=server.get('key_filename', None)) |
|
|
|
client.connect(server['ip'], port=server['port'], username=server['username'], password=server.get('password', None), key_filename=server.get('key_filename', None), timeout=interval*2.5) |
|
|
|
cmd = 'nvidia-smi --query-gpu=index,name,memory.total,memory.used,memory.free,utilization.gpu,utilization.memory,temperature.gpu --format=csv' |
|
|
|
|
|
|
|
re_try_count = 0 |
|
|
|
# 循环检测 |
|
|
|
while run_realtime: |
|
|
|
try: |
|
|
|
stdin, stdout, stderr = client.exec_command(cmd) |
|
|
|
stdin, stdout, stderr = client.exec_command(cmd, timeout=interval*2.5) |
|
|
|
output = stdout.read().decode() |
|
|
|
output = output.split('\n') |
|
|
|
start_idx = 0 |
|
|
@ -416,7 +419,9 @@ def keep_check_one(server: dict, shared_data_list: list, server_idx: int, interv |
|
|
|
# if data_list_lock.locked and locked: |
|
|
|
# data_list_lock.release() |
|
|
|
# print(e) |
|
|
|
shared_data_list[server_idx]['err_info'] = str(e) |
|
|
|
shared_data_list[server_idx]['err_info'] = f'retry:{re_try_count}, {e}' |
|
|
|
time.sleep(re_connect_time) |
|
|
|
re_try_count += 1 |
|
|
|
|
|
|
|
def realtime(args): |
|
|
|
global run_realtime |
|
|
|