hosting/clore_hosting/main.py

554 lines
28 KiB
Python

from lib import config as config_module
from lib import logging as logging_lib
from lib import log_streaming_task
from lib import run_startup_script
from lib import hive_miner_interface
from lib import docker_interface
from lib import docker_deploy
from lib import docker_pull
from lib import get_specs
from lib import utils
from lib import nvml
log = logging_lib.log
from clore_hosting import docker_configurator
from clore_hosting import api_interface
from clore_hosting import ws_interface
from clore_hosting import types
from queue import Queue
import concurrent.futures
import threading
import asyncio
import time
import json
from aiofiles import os as async_os
import os
specs = get_specs.Specs()
container_log_broken = asyncio.Queue()
config = config_module.config
WebSocketClient = ws_interface.WebSocketClient(log_message_broker=container_log_broken)
#print(config)
async def configure_networks(containers):
res = await asyncio.to_thread(docker_configurator.configure, containers)
try:
fin_res = types.DockerConfiguratorRes(validation_and_security=res[0], valid_containers=res[1], use_hive_flightsheet=res[2])
return fin_res
except Exception as e:
return False
async def deploy_containers(validated_containers, allowed_running_containers):
try:
all_running_container_names, all_stopped_container_names = await asyncio.to_thread(docker_deploy.deploy, validated_containers, allowed_running_containers)
return types.DeployContainersRes(all_running_container_names=all_running_container_names, all_stopped_container_names=all_stopped_container_names)
except Exception as e:
return False
async def get_local_images(no_latest_tag = False):
res = await asyncio.to_thread(docker_interface.get_local_images, no_latest_tag)
return res
async def set_oc(settings):
try:
result = await asyncio.to_thread(nvml.set_oc, settings)
return result
except Exception as e:
log.error(f"set_oc() | error | {e}")
return False
async def set_hive_miner_status(enabled=False):
try:
result = await asyncio.to_thread(utils.hive_set_miner_status, enabled)
return True
except Exception as e:
log.error(f"set_hive_miner_status() | error | {e}")
return False
class CloreClient:
def __init__(self, auth_key):
self.auth_key=auth_key
self.ws_peers = {}
self.last_checked_ws_peers=0
self.containers={}
self.needed_images=[]
self.containers_set=False
self.allowed_images = None
self.p_needed_containers=[]
self.last_pull_progress={}
self.validated_containers_set=False
self.validated_containers=[]
self.all_running_container_names=[]
self.all_stopped_container_names=[]
self.last_hw_specs_submit = time.time()-(1800-60)
self.last_service_heartbeat = {
"main": utils.unix_timestamp(),
"handle_container_cache": utils.unix_timestamp(),
"startup_script_runner": utils.unix_timestamp(),
"log_streaming_task": utils.unix_timestamp(),
"container_log_streaming_service": utils.unix_timestamp(),
"specs_service": utils.unix_timestamp(),
"oc_service": utils.unix_timestamp(),
"background_pow_data_collection": utils.unix_timestamp()
}
self.max_service_inactivity = 600 # seconds
if config.debug_ws_peer:
self.ws_peers[str(config.debug_ws_peer)]={
"expiration":"immune"
}
self.os_release = get_specs.get_os_release()
self.restart_docker = False
if "use_cgroupfs" in self.os_release:
self.updated_exec_opts = True if docker_interface.configure_exec_opts("native.cgroupdriver","cgroupfs") else False
if self.updated_exec_opts:
docker_info = docker_interface.get_info()
if "CgroupDriver" in docker_info and docker_info["CgroupDriver"]=="systemd":
self.restart_docker = True # Restart docker when it's loaded under systemd (accual restart will happen only if no orders running to not disrupt workload)
docker_interface.verify_docker_version()
self.dont_use_hive_binaries = True if 'DONT_USE_HIVE_BINARIES' in os.environ else False
nvml.init(allow_hive_binaries=not self.dont_use_hive_binaries)
self.extra_allowed_images = utils.get_extra_allowed_images()
self.allowed_running_containers = utils.get_allowed_container_names()
self.gpu_oc_specs = nvml.get_gpu_oc_specs()
self.last_oc_service_submit = 0
self.last_applied_oc = {}
self.last_oc_apply_time = 0
self.is_hive = get_specs.is_hive()
self.use_hive_flightsheet = False
self.hive_miner_interface = hive_miner_interface.hive_interface()
self.next_pow_background_job_send_update = 0
async def service(self):
global container_log_broken
pull_list = asyncio.Queue()
monitoring = asyncio.Queue()
task1 = asyncio.create_task(self.main(pull_list, monitoring))
task2 = asyncio.create_task(self.handle_container_cache(pull_list, monitoring))
task3 = asyncio.create_task(self.startup_script_runner(monitoring))
task4 = asyncio.create_task(log_streaming_task.log_streaming_task(container_log_broken, monitoring, self.allowed_running_containers))
task5 = asyncio.create_task(self.container_log_streaming_service(monitoring))
task6 = asyncio.create_task(self.specs_service(monitoring))
task7 = asyncio.create_task(self.oc_service(monitoring))
task8 = asyncio.create_task(self.background_pow_data_collection(monitoring))
monitoring_task = asyncio.create_task(self.monitoring_service(monitoring))
# Wait for both tasks to complete (they won't in this case)
await asyncio.gather(task1, task2, task3, task4, task5, task6, task7, task8, monitoring_task)
async def monitoring_service(self, monitoring):
while True:
try:
monitoring_data = []
while not monitoring.empty():
monitoring_data.append(await monitoring.get())
if len(monitoring_data)>0:
unique_monitoring = list(set(monitoring_data))
for service_name in unique_monitoring:
self.last_service_heartbeat[service_name]=utils.unix_timestamp()
if config.debug:
log.success(self.last_service_heartbeat)
for service_name in self.last_service_heartbeat.keys():
last_hearthbeat = self.last_service_heartbeat[service_name]
if last_hearthbeat < utils.unix_timestamp()-config.maximum_pull_service_loop_time and service_name=="handle_container_cache":
log.error(f"\"{service_name}\" service is stuck for {utils.unix_timestamp()-last_hearthbeat} s, Restarting...")
os._exit(1)
elif last_hearthbeat < utils.unix_timestamp()-config.maximum_service_loop_time and service_name!="handle_container_cache":
log.error(f"\"{service_name}\" service is stuck for {utils.unix_timestamp()-last_hearthbeat} s, Restarting...")
os._exit(1)
except Exception as e:
log.debug(f"monitoring_service() | ERROR | {e}")
await asyncio.sleep(5)
async def container_log_streaming_service(self, monitoring):
while True:
try:
await monitoring.put("container_log_streaming_service")
await WebSocketClient.stream_container_logs()
except Exception as e:
log.debug(f"container_log_streaming_service() | ERROR | {e}")
await asyncio.sleep(0.6)
async def run_startup_scripts(self, startup_script_full_path, container_name):
try:
if config.debug:
log.success(f"Runnin' {startup_script_full_path}")
log.error(self.all_running_container_names)
await asyncio.to_thread(run_startup_script.run, container_name, startup_script_full_path, f"/init-{container_name}.sh")
return True
except Exception as e:
return False
async def startup_script_runner(self, monitoring):
startup_script_ongoing_tasks = {}
while True:
try:
await monitoring.put("startup_script_runner")
startup_script_files = await async_os.listdir(config.startup_scripts_folder)
for startup_script_file in startup_script_files:
if type(startup_script_file)==str and startup_script_file.endswith(".sh") and startup_script_file[:-3] in self.all_running_container_names:
if not f"{startup_script_file[:-3]}.finished" in startup_script_files:
full_startup_script_path = os.path.join(config.startup_scripts_folder, startup_script_file)
if os.path.isfile(full_startup_script_path) and full_startup_script_path not in startup_script_ongoing_tasks:
# Start processing the file immediately in a non-blocking way
startup_script_task = asyncio.create_task(self.run_startup_scripts(full_startup_script_path, startup_script_file[:-3]))
startup_script_ongoing_tasks[full_startup_script_path] = startup_script_task
# Attach a callback to clean up the task once it's done
startup_script_task.add_done_callback(lambda t, path=full_startup_script_path: startup_script_ongoing_tasks.pop(path, None))
# Remove completed tasks
completed_tasks = [path for path, task in startup_script_ongoing_tasks.items() if task.done()]
for path in completed_tasks:
startup_script_ongoing_tasks.pop(path, None)
except Exception as e:
log.debug(f"ERROR | startup_script_runner() | {e}")
await asyncio.sleep(2)
async def pull_log_progress(self, log_dict, image_name):
while True:
tmp_progress=''
for layer, status in log_dict.items():
first_char = '' if tmp_progress=='' else '\n'
tmp_progress+=f"{first_char}{layer}: {status}"
self.last_pull_progress[image_name]={"log":tmp_progress, "last_update":time.time()}
await asyncio.sleep(config.pull_log_streaming_interval/1000)
async def check_if_pulling_required(self, pulling_image, cancellation_event):
while True:
try:
matched_image = False
for needed_image in self.needed_images:
if "image" in needed_image and needed_image["image"]==pulling_image:
matched_image=True
break
if not matched_image:
cancellation_event.set()
except Exception as e:
pass
await asyncio.sleep(0.5)
async def handle_container_cache(self, pull_list, monitoring):
while True:
got_data = []
while not pull_list.empty():
got_data.append(await pull_list.get())
await monitoring.put("handle_container_cache")
if len(got_data)>0:
self.p_needed_containers=got_data[len(got_data)-1]
if len(self.p_needed_containers)>0:
local_images = await get_local_images(no_latest_tag=True)
for local_image in local_images:
self.last_pull_progress[local_image]={"log":"Pull complete", "last_update":time.time()}
image_needed = False
removed_cnt = 0
try:
for p_needed_container in self.p_needed_containers:
if "image" in p_needed_container and local_image.replace(':latest','')==p_needed_container["image"].replace(':latest',''):
image_needed=True
break
if type(self.allowed_images)==list:
if image_needed==False:
after_split = local_image.split(':', 1)
local_image_name = after_split[0]
local_image_tag = ''
if len(after_split)>1:
local_image_tag=after_split[1]
for allowed_image in self.allowed_images:
if "repository" in allowed_image and "allowed_tags" in allowed_image and allowed_image["repository"]==local_image_name:
for allowed_tag in allowed_image["allowed_tags"]:
if local_image_tag==allowed_tag or allowed_tag=='*':
image_needed=True
break
if not image_needed and removed_cnt < config.max_remove_images_per_run and config.delete_unused_containers:
with concurrent.futures.ThreadPoolExecutor() as pool:
r = await asyncio.get_running_loop().run_in_executor(pool, docker_interface.remove_docker_image, local_image)
if r:
removed_cnt+=1
#if config.debug:
# log.success(f"{local_image} | {image_needed}")
except Exception as e:
image_needed=True
log.debug(f"ERROR | image_needed | {e}")
for lpp_image in self.last_pull_progress.keys():
log_info = self.last_pull_progress[lpp_image]
if log_info["last_update"] < time.time()-300:
del self.last_pull_progress[lpp_image]
break
most_recent_wanted_state = self.p_needed_containers
for wanted_image in most_recent_wanted_state:
if not wanted_image["image"] in local_images:
print("Local", local_images)
print("W",wanted_image)
log.debug(f"Starting to pull \"{wanted_image}\"")
auth_config = {}
if "dockerhub_token" in wanted_image and "dockerhub_user" in wanted_image:
auth_config={
"username": wanted_image["dockerhub_user"],
"password": wanted_image["dockerhub_token"]
}
log_dict = {}
pull_cancellation_event = asyncio.Event()
loop = asyncio.get_running_loop()
# Run the image pull, log progress concurrently and cancel if not needed anymore
pull_task = asyncio.create_task(docker_pull.pull_image(wanted_image["image"], auth_config, log_dict, loop, pull_cancellation_event))
log_task = asyncio.create_task(self.pull_log_progress(log_dict, wanted_image["image"]))
check_if_pulling_required_task = asyncio.create_task(self.check_if_pulling_required(wanted_image["image"], pull_cancellation_event))
# Wait for the image pull to complete, then cancel the log progress task
try:
await pull_task
except Exception as e:
self.last_pull_progress[local_image]={f"log":"Can't pull image \"{local_image}\"", "last_update":time.time()}
log_task.cancel()
try:
await log_task
except asyncio.CancelledError:
# Expect the task to be cancelled, so pass here
pass
check_if_pulling_required_task.cancel()
try:
await check_if_pulling_required_task
except asyncio.CancelledError:
# Expect the task to be cancelled, so pass here
pass
await asyncio.sleep(1)
async def main(self, pull_list, monitoring):
step=0
while True:
try:
step+=1
await monitoring.put("main")
if config.debug:
print("STEP",step,'|',self.containers_set, self.containers if config.log_containers_strings else '')
tasks = []
running_order = False
container_conf = WebSocketClient.get_containers()
if container_conf[0]:
self.containers_set=True
self.containers=container_conf[1]
tmp_images = []
for container in self.containers:
if "image" in container and "image" in container and container["image"]!="cloreai/hive-use-flightsheet":
log_pull = False
if "name" in container:
if "-order-" in container["name"]:
running_order=True
log_pull=True
image_config = {
"image":container["image"],
"log":log_pull
}
if "ip" in container and "| docker login -u " in container["ip"] and container["ip"][:8]=="; echo '":
try:
dockerhub_token = container["ip"][8:].split("'")[0]
dockerhub_user = container["ip"].split('docker login -u ')[1].split(';')[0][:-17]
image_config["dockerhub_token"]=dockerhub_token
image_config["dockerhub_user"]=dockerhub_user
except Exception as e:
log.error(e)
pass
if not image_config in tmp_images:
tmp_images.append(image_config)
if self.restart_docker and not running_order and len(self.containers)>0:
log.debug("Sending docker restart command")
utils.run_command_v2("systemctl restart docker")
self.restart_docker=False
if tmp_images!=self.needed_images:
self.needed_images=tmp_images
await pull_list.put(self.needed_images)
#self.containers.append({'name': 'clore-test', 'image': 'cloreai/monitoring:0.2', 'command': '', 'env': {'TOKEN': '22'}, 'gpus': True, 'network': 'clore-br69', 'ip': '172.22.0.23', 'network_subnet':'172.22.0.0/24', 'network_gateway':'172.22.0.1'})
if (self.last_checked_ws_peers < (utils.unix_timestamp()-config.ws_peers_recheck_interval)):
tasks.append(api_interface.get_server_config())
if self.containers_set:
tasks.append(configure_networks(self.containers))
tasks.append(WebSocketClient.stream_pull_logs())
if self.validated_containers_set:
tasks.append(deploy_containers(self.validated_containers, self.allowed_running_containers))
if step==1:
WebSocketClient.set_auth(self.auth_key)
asyncio.create_task(WebSocketClient.run())
elif step%5 == 0 and WebSocketClient.get_last_heartbeat() < (utils.unix_timestamp()-config.max_ws_peer_heartbeat_interval):
log.error(f"CLORE HOSTING | Didn't received heartbeat from clore.ai for over {config.max_ws_peer_heartbeat_interval} seconds")
log.error("CLORE HOSTING | exiting ...")
os._exit(1)
self.expire_ws_peers()
WebSocketClient.set_ws_peers(self.ws_peers)
WebSocketClient.set_pull_logs(self.last_pull_progress)
if len(tasks)>0:
results = await asyncio.gather(*tasks)
# Process the results (optional)
for result in results:
if type(result)==types.ServerConfig:
if result.success:
self.last_checked_ws_peers = utils.unix_timestamp()
self.allowed_images=result.allowed_images+self.extra_allowed_images
if not config.debug_ws_peer:
for pure_ws_peer in result.ws_peers:
self.ws_peers[pure_ws_peer]={
"expiration":utils.unix_timestamp()+900
}
elif self.allowed_images==None:
log.error("Can't contact clore.ai, restarting")
os._exit(1)
elif type(result)==types.DockerConfiguratorRes:
if result.validation_and_security:
self.validated_containers_set=True
self.validated_containers = result.valid_containers
self.use_hive_flightsheet = result.use_hive_flightsheet
log.debug(f"Use Hive flightsheet: {result.use_hive_flightsheet}")
elif type(result)==types.DeployContainersRes:
try:
self.all_running_container_names = result.all_running_container_names
self.all_stopped_container_names = result.all_stopped_container_names
except Exception as e:
pass
except Exception as e:
log.debug(f"main() | ERROR | {e}")
await asyncio.sleep(1)
async def submit_specs(self, current_specs):
try:
if type(current_specs) == dict:
current_specs["backend_version"]=16
current_specs["update_hw"]=True
smallest_pcie_width = 999
for gpu in current_specs["gpus"]["nvidia"]:
if "pcie_width" in gpu and gpu["pcie_width"]<smallest_pcie_width:
smallest_pcie_width=gpu["pcie_width"]
current_specs["pcie_width"]=gpu["pcie_width"]
current_specs["pcie_rev"]=gpu["pcie_revision"]
await WebSocketClient.send(current_specs)
except Exception as e:
log.debug(f"FAIL | submit_specs() | {e}")
async def update_realtime_data(self, current_specs):
try:
if type(current_specs) == dict:
cpu_usage = await get_specs.get_cpu_usage()
ram_usage = await get_specs.get_ram_usage()
gpu_list = current_specs["gpus"]["nvidia"]+current_specs["gpus"]["amd"]
submit_document = {
"update_realtime_data":True,
"gpus": gpu_list,
"cpu": cpu_usage,
"ram": ram_usage,
"all_running_container_names": self.all_running_container_names,
"all_stopped_container_names": self.all_stopped_container_names
}
await WebSocketClient.send(submit_document)
except Exception as e:
log.debug(f"FAIL | update_realtime_data() | {e}")
async def specs_service(self, monitoring):
while True:
try:
await monitoring.put("specs_service")
current_specs = await specs.get()
if self.last_hw_specs_submit < (utils.unix_timestamp()-1800):
self.last_hw_specs_submit=utils.unix_timestamp()
await self.submit_specs(current_specs)
await self.update_realtime_data(current_specs)
except Exception as e:
log.debug(f"FAIL | specs_service() | {e}")
await asyncio.sleep(7)
async def oc_service(self, monitoring):
while True:
try:
await monitoring.put("oc_service")
oc_apply_allowed = True
### OC Service should also hande Hive stuff
if self.use_hive_flightsheet and self.is_hive and not self.dont_use_hive_binaries:
await set_hive_miner_status(True)
oc_apply_allowed = False # Don't apply any OC when running HiveOS miner
elif self.is_hive and not self.dont_use_hive_binaries:
await set_hive_miner_status(False)
### Run OC tasks
oc_conf = WebSocketClient.get_oc()
if oc_conf[0] and type(self.gpu_oc_specs)==list and oc_conf[1]!=self.gpu_oc_specs and self.last_oc_service_submit+240 < utils.unix_timestamp():
log.debug("Submitting \"gpu_oc_specs\"")
self.last_oc_service_submit = utils.unix_timestamp()
await WebSocketClient.send({
"set_gpu_info":self.gpu_oc_specs,
"xorg_valid": True
})
if oc_conf[0] and type(oc_conf[2])==dict and oc_apply_allowed:
if utils.normalize_rule(self.last_applied_oc)!=utils.normalize_rule(oc_conf[2]) or (self.last_oc_apply_time < utils.unix_timestamp()-300):
self.last_oc_apply_time = utils.unix_timestamp()
log.debug(f"Applying OC | {json.dumps(oc_conf[2], separators=(',',':'))}")
await set_oc(oc_conf[2])
self.last_applied_oc=oc_conf[2]
except Exception as e:
log.debug(f"FAIL | oc_service() | {e}")
await asyncio.sleep(2)
async def background_pow_data_collection(self, monitoring):
while True:
try:
await monitoring.put("background_pow_data_collection")
if not self.dont_use_hive_binaries and self.is_hive:
miner_config = await self.hive_miner_interface.export_miner_stats(get_hashrates=False)
if (miner_config["miner_uptime"]>0 and miner_config["miner_uptime"]<60) or self.next_pow_background_job_send_update < time.time():
self.next_pow_background_job_send_update = time.time()+(5*60)
current_statistics = await self.hive_miner_interface.export_miner_stats(get_hashrates=True)
submit_result = await WebSocketClient.send({"submit_hashrates": current_statistics})
if not submit_result:
self.next_pow_background_job_send_update = time.time()+40
except Exception as e:
log.debug(f"FAIL | background_pow_data_collection() | {e}")
await asyncio.sleep(6)
def expire_ws_peers(self):
for ws_peer_address in list(self.ws_peers.keys()):
ws_peer_info = self.ws_peers[ws_peer_address]
try:
if ws_peer_info["expiration"]!="immune" and (ws_peer_info["expiration"] < utils.unix_timestamp()):
self.ws_peers.pop(ws_peer_address, None)
except Exception as e:
pass