Looking at the logs in the Kube pods now for anything that looks unusual...
Can this be reproducible using a simple script that we can also run?
Not really unfortunately - happy to share my code, but I've managed to reproduce this with different codebases.
As a summary of what I've tried:
- Agent on the H100 machine, Server on Kube - Fail
- Agent on laptop, Server on Kube - Fail
- Agent on laptop, Server on Docker Desktop - Pass
So I'm 100% sure there is something wrong with our ClearML Server deployment on Kube rather than an issue with the agents or code. As for which of the 7 containers could be at fault... :man-shrugging: . I'm not seeing anything out of the ordinary in the logs. Is there a verbose setting in the agent that could help us diagnose, i.e. each step of what goes on inTask.init?
Retrying (Retry(total=239, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
OH that makes sense I'm assuming on your local machine the certificate is installed but not on remote machines / containers
Add the following to your clearml.conf:
api.verify_certificate: false
- Agent on laptop, Server on Kube - Fail
So I'm 100% sure there is something wrong with our ClearML Server deployment on Kube
Yeah that feels like a network config issue...
Is there a verbose setting in the agent that could help us diagnose,
yes running with debug turned on on.
since you managed to reproduce on your latop you can try to run the agent with --debug to test, specifically:
clearml-agent --debug daemon ....
if you are running it in venv mode (which I think the setup) you can also just specify the Task ID and test that (no daemon just execution)
clearml-agent --debug execute --id <task_id_here>
My understanding is that on remote execution Task.init is supposed to be a no-op right?
Not really a no-op, it would sync Argpasrer and the like, start background reporting services etc.
This is so odd! literally nothing printed
Can you tell me something about the node "mrl-plswh100:0" ?
is this like a sagemaker node? we have seen things similar where Python threads / subprocesses are not supported and instead of python crashing it just hangs there
Our server is deployed on a kube cluster. I'm not too clear on how Helm charts etc.
The only thing that I can think of is that something is not right the the load balancer on the server so maybe some requests coming from an instance on the cluster are blocked ...
Hmm, saying that aloud that actually could be?! Try to add the following line to the end of the clearml.conf on the machine running the agent:
api.http.default_method: "put"
confirmed that the change had been added by
Make sure you see them in the Task log in the UI (the agent print it when it starts)
Any insight on how we can reproduce the issue?
Can this be reproducible using a simple script that we can also run?
Hi @<1523701205467926528:profile|AgitatedDove14> , I reordered the imports:
from clearml import Task
print("Before task")
task = Task.init(project_name="ClearML Testing", task_name="FMNIST")
task.set_repo(
repo="git@ssh.dev.azure.com:v3/mclarenracing/Application%20Engineering/ml-queue-test"
)
task.set_packages("requirements.txt")
print("After task")
print("Before import")
from pathlib import Path
import hydra
import lightning as L
import torch
from coolname import generate_slug
from omegaconf import DictConfig
from src.datasets import JobDataModule
from src.models import JobModel
from src.utils import LogSummaryCallback, get_num_steps, prepare_loggers_and_callbacks
for i in range(torch.cuda.device_count()):
print(torch.cuda.get_device_properties(i).name)
And here's the output:
Environment setup completed successfully
Starting Task Execution:
Before task
Still looks like it's getting stuck at Task.init
Thanks for the response @<1523701205467926528:profile|AgitatedDove14> ! The code is a small FMNIST test training job written in PyTorch Lightning. On my local job (laptop GPU, Windows) it completes in ~ 5min. On the server (Linux, H100s) it just hangs at Starting Task Execution: . Neither of these are in Docker.
I would expect to see the standard PL progress bars outputted to the console, but since nothing is outputted, so I'm not sure how to go about debugging this. I've attached the full logs for local and remote
Thanks Martin - will try that and see what I can find. Really appreciate your patience with this! 🙂
If there was an SSL issue it should log to console right?
ClearML is hosted on an on-prem kube cluster and to get it to log locally I needed to append my company cert to the file located at certifi.where() . Do you think the same needs to be done for the Python installation for the worker?
This is so odd,
could you add prints right after the Task.init?
Also could you verify it still gets stuck with the latest RC
clearml==1.16.3rc2
Although it's still really weird how it was failing silently
totally agree, I think the main issue was the agent had the correct configuration, but the container / env the agent was spinning was missing it,
I'll double check how come it did not print anything
@<1523701205467926528:profile|AgitatedDove14> we've now configured the server to have it's own user account to run the agent so it is no longer running as root, but no luck 😞
Before os.environ
environ({'LANG': 'en_GB.UTF-8', 'PATH': '/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin', 'HOME': '/home/clearml', 'LOGNAME': 'clearml', 'USER': 'clearml', 'SHELL': '/bin/bash', 'INVOCATION_ID': 'da8e36a03c7348efbb7db360755e92b3', 'JOURNAL_STREAM': '8:244189055', 'SYSTEMD_EXEC_PID': '1970812', 'PYTHONUNBUFFERED': '1', 'CUDA_DEVICE_ORDER': 'PCI_BUS_ID', 'CLEARML_WORKER_ID': 'mrl-plswh100:0', 'TRAINS_WORKER_ID': 'mrl-plswh100:0', 'CLEARML_CONFIG_FILE': '/tmp/.clearml_agent.4ll2u471.cfg', 'TRAINS_CONFIG_FILE': '/tmp/.clearml_agent.4ll2u471.cfg', 'CLEARML_TASK_ID': '4ab4c22b02ed4d1f86ff4fac663828f0', 'TRAINS_TASK_ID': '4ab4c22b02ed4d1f86ff4fac663828f0', 'CLEARML_LOG_LEVEL': 'INFO', 'TRAINS_LOG_LEVEL': 'INFO', 'CLEARML_LOG_TASK_TO_BACKEND': '0', 'TRAINS_LOG_TASK_TO_BACKEND': '0', 'PYTHONPATH': '/home/clearml/.clearml/venvs-builds/3.9/task_repository/ml-queue-test:/home/clearml/.clearml/venvs-builds/3.9/task_repository/ml-queue-test::/usr/lib64/python39.zip:/usr/lib64/python3.9:/usr/lib64/python3.9/lib-dynload:/home/clearml/.clearml/venvs-builds/3.9/lib64/python3.9/site-packages:/home/clearml/.clearml/venvs-builds/3.9/lib/python3.9/site-packages'})
Before Task.init
This is exactly my problem, too, which I described above! If you find any solution, would be glad if you could share. 🙂 Of course, I also share mine when I get one.
@<1724960464275771392:profile|DepravedBee82> I just realized, the agent is Not running in docker mode, correct? (i.e. venv mode)
If this is the case how come it is running as root? (could it be is is running inside a container? how was that container spinned?)
Ok so my train.py now looks like this:
print("Before import")
from pathlib import Path
import hydra
import lightning as L
import torch
from coolname import generate_slug
from omegaconf import DictConfig
from src.datasets import JobDataModule
from src.models import JobModel
from src.utils import LogSummaryCallback, get_num_steps, prepare_loggers_and_callbacks
from clearml import Task
for i in range(torch.cuda.device_count()):
print(torch.cuda.get_device_properties(i).name)
print("Before task")
task = Task.init(project_name="ClearML Testing", task_name="FMNIST")
task.set_repo(
repo="git@ssh.dev.azure.com:v3/mclarenracing/Application%20Engineering/ml-queue-test"
)
task.set_packages("requirements.txt")
print("After task")
And the log looks like this:
Starting Task Execution:
Before import
2024-07-19 09:06:09
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
NVIDIA H100 80GB HBM3
Before task
So it looks like it's getting stuck at Task.init
Will try non-root and get back to you. I’m also trying to reproduce on a different machine too
My money is on the Redis container although comparing the logs between Kube & Docker Desktop, nothing looks out of the ordinary...
Okay I have an idea, it could be a lock that another agent/user is holding on the cache folder or similar
Let me check something
Sorry, on the remote machine (i.e. enqueue it and let the agent run it), this will also log the print 🙂
My understanding is that on remote execution Task.init is supposed to be a no-op right?
Here's what the agent was logging:
anjum.sayed@M209886 clearml-agent --debug daemon --queue default
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): api.clearml.dev.mrl:443
DEBUG:urllib3.connectionpool:
"PUT /auth.login HTTP/1.1" 200 603
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): api.clearml.dev.mrl:443
DEBUG:urllib3.connectionpool:
"PUT /v2.5/queues.get_all HTTP/1.1" 200 344
DEBUG:urllib3.connectionpool:
"PUT /v2.5/queues.get_all HTTP/1.1" 200 332
DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): updates.clear.ml:443
DEBUG:clearml_agent.session:Run by interpreter: C:\Users\anjum.sayed\AppData\Local\Programs\Python\Python39\python.exe
Current configuration (clearml_agent v1.8.1, location: C:\Users\anjum.sayed/clearml.conf):
----------------------
agent.worker_id =
agent.worker_name = M209886
agent.force_git_ssh_protocol = true
agent.python_binary =
agent.package_manager.type = pip
agent.package_manager.pip_version.0 = <20.2 ; python_version < '3.10'
agent.package_manager.pip_version.1 = <22.3 ; python_version >\= '3.10'
agent.package_manager.system_site_packages = false
agent.package_manager.force_upgrade = false
agent.package_manager.conda_channels.0 = pytorch
agent.package_manager.conda_channels.1 = conda-forge
agent.package_manager.conda_channels.2 = nvidia
agent.package_manager.conda_channels.3 = defaults
agent.package_manager.priority_optional_packages.0 = pygobject
agent.package_manager.torch_nightly = false
agent.package_manager.poetry_files_from_repo_working_dir = false
agent.venvs_dir = C:/Users/anjum.sayed/.clearml/venvs-builds
agent.venvs_cache.max_entries = 10
agent.venvs_cache.free_space_threshold_gb = 2.0
agent.venvs_cache.path = ~/.clearml/venvs-cache
agent.vcs_cache.enabled = true
agent.vcs_cache.path = C:/Users/anjum.sayed/.clearml/vcs-cache
agent.venv_update.enabled = false
agent.pip_download_cache.enabled = true
agent.pip_download_cache.path = C:/Users/anjum.sayed/.clearml/pip-download-cache
agent.translate_ssh = true
agent.reload_config = false
agent.docker_pip_cache = C:/Users/anjum.sayed/.clearml/pip-cache
agent.docker_apt_cache = C:/Users/anjum.sayed/.clearml/apt-cache
agent.docker_force_pull = false
agent.default_docker.image = nvidia/cuda:11.0.3-cudnn8-runtime-ubuntu20.04
agent.enable_task_env = false
agent.sanitize_config_printout = ****
agent.hide_docker_command_env_vars.enabled = true
agent.hide_docker_command_env_vars.parse_embedded_urls = true
agent.abort_callback_max_timeout = 1800
agent.docker_internal_mounts.sdk_cache = /clearml_agent_cache
agent.docker_internal_mounts.apt_cache = /var/cache/apt/archives
agent.docker_internal_mounts.ssh_folder = ~/.ssh
agent.docker_internal_mounts.ssh_ro_folder = /.ssh
agent.docker_internal_mounts.pip_cache = /root/.cache/pip
agent.docker_internal_mounts.poetry_cache = /root/.cache/pypoetry
agent.docker_internal_mounts.vcs_cache = /root/.clearml/vcs-cache
agent.docker_internal_mounts.venv_build = ~/.clearml/venvs-builds
agent.docker_internal_mounts.pip_download = /root/.clearml/pip-download-cache
agent.apply_environment = true
agent.apply_files = true
agent.custom_build_script =
agent.disable_task_docker_override = false
agent.git_user =
agent.git_pass = ****
agent.git_host =
agent.debug = true
agent.default_python = 3.9
agent.cuda_version = 123
agent.cudnn_version = 0
api.version = 1.5
api.verify_certificate = true
api.default_version = 1.5
api.http.max_req_size = 15728640
api.http.retries.total = 240
api.http.retries.connect = 240
api.http.retries.read = 240
api.http.retries.redirect = 240
api.http.retries.status = 240
api.http.retries.backoff_factor = 1.0
api.http.retries.backoff_max = 120.0
api.http.wait_on_maintenance_forever = true
api.http.pool_maxsize = 512
api.http.pool_connections = 512
api.http.default_method = put
api.auth.token_expiration_threshold_sec = ****
api.api_server =
api.web_server =
api.files_server =
api.credentials.access_key = 1N33K4IXUYO64HVT4S3PXVDIX4K2CS
api.credentials.secret_key = ****
api.host =
sdk.storage.cache.default_base_dir = ~/.clearml/cache
sdk.storage.cache.size.min_free_bytes = 10GB
sdk.storage.direct_access.0.url = file://*
sdk.metrics.file_history_size = 100
sdk.metrics.matplotlib_untitled_history_size = 100
sdk.metrics.images.format = JPEG
sdk.metrics.images.quality = 87
sdk.metrics.images.subsampling = 0
sdk.metrics.tensorboard_single_series_per_graph = false
sdk.network.metrics.file_upload_threads = 4
sdk.network.metrics.file_upload_starvation_warning_sec = 120
sdk.network.iteration.max_retries_on_server_error = 5
sdk.network.iteration.retry_backoff_factor_sec = 10
sdk.network.file_upload_retries = 3
sdk.aws.s3.key =
sdk.aws.s3.secret = ****
sdk.aws.s3.region =
sdk.aws.s3.use_credentials_chain = false
sdk.aws.boto3.pool_connections = 512
sdk.aws.boto3.max_multipart_concurrency = 16
sdk.aws.boto3.multipart_threshold = 8388608
sdk.aws.boto3.multipart_chunksize = 8388608
sdk.log.null_log_propagate = false
sdk.log.task_log_buffer_capacity = 66
sdk.log.disable_urllib3_info = true
sdk.development.task_reuse_time_window_in_hours = 72.0
sdk.development.vcs_repo_detect_async = true
sdk.development.store_uncommitted_code_diff = true
sdk.development.support_stopping = true
sdk.development.default_output_uri =
sdk.development.force_analyze_entire_repo = false
sdk.development.suppress_update_message = false
sdk.development.detect_with_pip_freeze = false
sdk.development.worker.report_period_sec = 2
sdk.development.worker.ping_period_sec = 30
sdk.development.worker.log_stdout = true
sdk.development.worker.report_global_mem_used = false
sdk.development.worker.report_event_flush_threshold = 100
sdk.development.worker.console_cr_flush_period = 10
sdk.apply_environment = false
sdk.apply_files = false
DEBUG:clearml_agent.commands.worker:starting resource monitor thread
Worker "M209886:0" - Listening to queues:
+----------------------------------+---------+-------+
| id | name | tags |
+----------------------------------+---------+-------+
| 3e9973e15a6048c5ae5419ea7d097f9c | default | |
+----------------------------------+---------+-------+
DEBUG:urllib3.connectionpool:
"PUT /workers.register HTTP/1.1" 200 278
Running CLEARML-AGENT daemon in background mode, writing stdout/stderr to C:\Users\ANJUM~1.SAY\AppData\Local\Temp\.clearml_agent_daemon_outg5aq488v.txt
DEBUG:urllib3.connectionpool:
"PUT /v2.5/queues.get_all HTTP/1.1" 200 337
DEBUG:urllib3.connectionpool:
"PUT /workers.get_runtime_properties HTTP/1.1" 404 371
DEBUG:urllib3.connectionpool:
"PUT /v2.14/queues.get_next_task HTTP/1.1" 200 282
.................. truncating due to Slack char limit.........
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.5/tasks.ping HTTP/1.1" 200 271
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"POST /events.add_batch HTTP/1.1" 200 315
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /workers.status_report HTTP/1.1" 200 283
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /workers.status_report HTTP/1.1" 200 283
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.5/tasks.ping HTTP/1.1" 200 271
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.14/tasks.get_all HTTP/1.1" 200 363
DEBUG:urllib3.connectionpool:
"PUT /v2.5/tasks.get_by_id HTTP/1.1" 200 3490
DEBUG:urllib3.connectionpool:
"PUT /v2.5/tasks.stopped HTTP/1.1" 200 304
INFO:clearml_agent.commands.worker:Task process terminated
INFO:clearml_agent.commands.worker:Task interrupted: stopping
DEBUG:urllib3.connectionpool:
"POST /events.add_batch HTTP/1.1" 200 315
DEBUG:urllib3.connectionpool:
"PUT /v2.5/tasks.stopped HTTP/1.1" 200 333
DEBUG:urllib3.connectionpool:
"PUT /workers.status_report HTTP/1.1" 200 283
DEBUG:urllib3.connectionpool:
"PUT /v2.5/queues.get_all HTTP/1.1" 200 337
DEBUG:urllib3.connectionpool:
"PUT /v2.14/queues.get_next_task HTTP/1.1" 200 282
DEBUG:urllib3.connectionpool:
"PUT /workers.unregister HTTP/1.1" 200 280
DEBUG:urllib3.connectionpool:
"PUT /workers.unregister HTTP/1.1" 200 280
I think I've found a clue after running with debug:
Before Task.init
Retrying (Retry(total=239, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
Retrying (Retry(total=238, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
2024-07-30 10:20:07
Retrying (Retry(total=237, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
2024-07-30 10:20:12
Retrying (Retry(total=236, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
2024-07-30 10:20:33
Retrying (Retry(total=235, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
2024-07-30 10:21:03
Retrying (Retry(total=234, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
2024-07-30 10:22:08
Retrying (Retry(total=233, connect=240, read=240, redirect=240, status=240)) after connection broken by 'SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1129)'))': /auth.login
So my current theory is that the env the agent is building doesn't have the corporate TLS/SSL certificates. It's weird how it was failing silently without the --debug flag though...
It’s a Dell XE9680 rack server with 8xH100s which is located in our office, running AlmaOS. We have successfully run training jobs on it inside Docker (without ClearML) which work fine (will check with my team if we’ve got something to train without Docker). I’ve also tried different Python versions; 3.9 (Alma default) and 3.11 which you can see in the log above. It’s a really bizarre issue and outside of print statements I’m not really sure where to look.
You mentioned sync argparser & reporting, so I’ll try removing Hydra to rule that out, and other loggers in PL and see from there …
I managed to set up my (Windows) laptop as a worker and reproduce the issue.
Any insight on how we can reproduce the issue?
Nope - confirmed to be running on the OS's Python environment, although he said that the agent was supposed to have it's own user - looking into that now
I just ran with this in my local task, and all the env vars were printed to console, but in ClearML they are not in the console log. Presumably that's because it's printed before ClearML is logging?