bowong
/
modalDeploy
6
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

FIX autodl自动创建修复

This commit is contained in:
kyj@bowong.ai 2025-04-15 11:11:34 +08:00
parent 43479f043d
commit 61106608e0
1 changed files with 17 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
AutoDL/audodl_sdk.py
View File

@ -11,6 +11,7 @@ req_instance_page_size = 1500
chk_instance_page_size = 1500 chk_instance_page_size = 1500
instances = {} instances = {}
LIM = 30 # 等待状态时间LIM*5s LIM = 30 # 等待状态时间LIM*5s
START_LIM = 20 #Heygem脚本启动等待超时时间-10
def ssh_try(host,port,pwd): def ssh_try(host,port,pwd):
# 建立连接 # 建立连接
@ -21,10 +22,22 @@ def ssh_try(host,port,pwd):
ssh = paramiko.SSHClient() ssh = paramiko.SSHClient()
ssh._transport = trans ssh._transport = trans
# 剩下的就和上面一样了
ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy()) ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())
ssh_stdin, ssh_stdout, ssh_stderr = ssh.exec_command("/bin/bash -lc \"source /root/.bashrc && nohup python /root/AutoDL_pure_heygem.py > log.txt 2>&1 &\"") ssh_stdin, ssh_stdout, ssh_stderr = ssh.exec_command("bash -ic \"source /root/.bashrc && sleep 10 && ps -ef|grep heygem|grep -v grep && lsof -i:6006|wc -l\"", get_pty=True)
out = ssh_stdout.read().decode()
start_lim = START_LIM
if len(out.split("\n")[-2]) > 2:
trans.close()
raise RuntimeError(out.split("\n")[-2])
while int(out.split("\n")[-2]) <= 1 and start_lim > 0:
ssh_stdin, ssh_stdout, ssh_stderr = ssh.exec_command("bash -ic \"sleep 2 && lsof -i:6006|wc -l\"")
out = ssh_stdout.read().decode()
loguru.logger.info("waiting for HeyGem server ready...")
start_lim -= 1
if start_lim <= 0:
loguru.logger.error("HeyGem Server Start Timeout, Please check!")
else:
loguru.logger.success("HeyGem Server Start Success")
# 关闭连接 # 关闭连接
trans.close() trans.close()
@ -239,7 +252,7 @@ if __name__=="__main__":
machines = get_autodl_machines() machines = get_autodl_machines()
for m in machines: for m in machines:
instance_uuid, domain = payg(m["region_name"], m["machine_id"]) instance_uuid, domain = payg(m["region_name"], m["machine_id"])
print(instance_uuid, "https://"+domain) loguru.logger.success("instance_id:%s server_domain: https://%s" % (instance_uuid, domain))
# if instance_uuid: # if instance_uuid:
# instance_operate(instance_uuid, "power_off") # instance_operate(instance_uuid, "power_off")
# instance_operate(instance_uuid, "release") # instance_operate(instance_uuid, "release")