clore
/
hosting
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

dev: debug

This commit is contained in:
empresa 2025-10-16 00:49:59 +07:00
parent daa634bfa9
commit 96bd92e5f0
2 changed files with 3 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
clore_hosting/main.py
View File

@ -554,10 +554,13 @@ class CloreClient:
try: try:
await monitoring.put("specs_service") await monitoring.put("specs_service")
current_specs = await specs.get() current_specs = await specs.get()
print(current_specs)
if self.last_hw_specs_submit < (utils.unix_timestamp()-1800): if self.last_hw_specs_submit < (utils.unix_timestamp()-1800):
self.last_hw_specs_submit=utils.unix_timestamp() self.last_hw_specs_submit=utils.unix_timestamp()
await self.submit_specs(current_specs) await self.submit_specs(current_specs)
print("submit specs")
await self.update_realtime_data(current_specs) await self.update_realtime_data(current_specs)
print("update realtime")
try: try:
if self.xfs_state == "active" and len(current_specs["gpus"]["nvidia"]) > 0 and not self.runned_pull_selftest: if self.xfs_state == "active" and len(current_specs["gpus"]["nvidia"]) > 0 and not self.runned_pull_selftest:
await clore_partner.check_to_pull_selftest(current_specs) await clore_partner.check_to_pull_selftest(current_specs)

12
lib/get_specs.py
View File

@ -291,8 +291,6 @@ def get_gpu_info():
"mem_used": parts[12] "mem_used": parts[12]
} }
print(xl_gpu_info)
try: try:
pci_query = parts[2][parts[2].find(':')+1:] pci_query = parts[2][parts[2].find(':')+1:]
for index, valid_pci_dev in enumerate(valid_pci_dev_list): for index, valid_pci_dev in enumerate(valid_pci_dev_list):
@ -302,8 +300,6 @@ def get_gpu_info():
xl_gpu_info["pcie_width"]=bus_spec.width xl_gpu_info["pcie_width"]=bus_spec.width
xl_gpu_info["pcie_revision"]=bus_spec.revision xl_gpu_info["pcie_revision"]=bus_spec.revision
except Exception as e: except Exception as e:
print("err")
print(e)
pass pass
gpus["nvidia"].append(xl_gpu_info) gpus["nvidia"].append(xl_gpu_info)
lines = nvidia_smi_stdout.split('\n') lines = nvidia_smi_stdout.split('\n')
@ -313,21 +309,13 @@ def get_gpu_info():
gpu_name = parts[1].strip() gpu_name = parts[1].strip()
gpu_id = parts[5].strip(); gpu_id = parts[5].strip();
print("parts")
print(parts)
print("p1:" + gpu_name)
print("p1:" + gpu_id)
if gpu_name == "NVIDIA Graphics Device" and gpu_id in GPU_ID_TO_NAME: if gpu_name == "NVIDIA Graphics Device" and gpu_id in GPU_ID_TO_NAME:
gpu_name = GPU_ID_TO_NAME[gpu_id] gpu_name = GPU_ID_TO_NAME[gpu_id]
print("p2:" + gpu_name)
gpu_str = f"{len(lines)-1}x {gpu_name}" gpu_str = f"{len(lines)-1}x {gpu_name}"
print(gpu_str)
gpu_mem = round(int(filter_non_numeric(parts[4]).strip())/1024, 2) gpu_mem = round(int(filter_non_numeric(parts[4]).strip())/1024, 2)
except Exception as e: except Exception as e:
nvml_err=True nvml_err=True
print("err2")
print(e)
pass pass
else: else:
nvml_err=True nvml_err=True