SageMaker SSH Helper

SageMaker SSH Helper is a library that helps you to securely connect to Amazon SageMaker's training jobs, processing jobs, realtime inference endpoints, and SageMaker Studio notebook containers for fast interactive experimentation, remote debugging, and advanced troubleshooting.

The two most common scenarios for the library, also known as "SSH into SageMaker", are:

Open a terminal session into a container running in SageMaker to diagnose a stuck training job, use CLI commands like nvidia-smi, or iteratively fix and re-execute your training script within seconds.
Remote debug a code running in SageMaker from your local favorite IDE like PyCharm Professional Edition or Visual Studio Code.

Other scenarios include but not limited to connecting to a remote Jupyter Notebook in SageMaker Studio from your IDE, connect with your browser to a TensorBoard process running in the cloud, or start a VNC session to SageMaker Studio to run GUI apps.

Also see our Frequently Asked Questions, especially if you're using Windows on your local machine.

How it works

SageMaker SSH helper uses AWS Systems Manager (SSM) Session Manager, to register the SageMaker container in SSM, followed by creating an SSM session between your client machine and the SageMaker container. Then you can SSH into SageMaker by creating an SSH connection on top of the SSM session, that allows opening a Linux shell, and/or configuring bidirectional SSH port forwarding to enable applications like remote development/debugging/desktop, and others.

See detailed architecture diagrams of the complete flow of participating components in Training Diagram, and IDE integration with SageMaker Studio diagram.

Getting started

To get started, your AWS system administrator must set up needed IAM and SSM configuration in your AWS account as shown in Setting up your AWS account with IAM and SSM configuration.

Note: This solution is a sample AWS content. You should not use this content in your production accounts, in a production environment or on production or other critical data. If you plan to use the solution in production, please, carefully review it with your security team. You are responsible for testing, securing, and optimizing the sample content as appropriate for production grade use based on your specific business requirements, including any quality control practices and standards.

Use Cases

SageMaker SSH Helper supports a variety of use cases:

Connecting to SageMaker training jobs with SSM - open a shell to a single- or multi-node training job to examine its file systems, monitor resources, produce thread-dumps for stuck jobs, and interactively run your train script
Connecting to SageMaker inference endpoints with SSM
Connecting to SageMaker batch transform jobs
Connecting to SageMaker processing jobs
Remote debugging with PyCharm Debug Server over SSH - let SageMaker run your code that connects to PyCharm, to start line-by-line debugging with PyDev.Debugger, a.k.a. pydevd
Remote code execution with PyCharm / VSCode over SSH - let PyCharm run or debug your code line-by-line inside SageMaker container with SSH interpreter
Local IDE integration with SageMaker Studio over SSH for PyCharm / VSCode - iterate fast on a single node at early stages of development without submitting SageMaker jobs

If you want to add a new use case or a feature, see CONTRIBUTING.

Connecting to SageMaker training jobs with SSM

Download Demo (.mov)

Step 1: Install the library

Before starting the whole procedure, check that both pip and python commands point to Python version 3.7 or higher with python --version command.

Install the latest stable version of library from the PyPI repository:

pip install sagemaker-ssh-helper

Step 2: Modify your start training job code

Add import for SSHEstimatorWrapper
Add a dependencies parameter to the Estimator object.
Add an SSHEstimatorWrapper.create(estimator,...) call before calling fit() and add SageMaker SSH Helper as dependencies.
Add a call to ssh_wrapper.get_instance_ids() to get the SSM instance(s) id. We'll use this as the target to connect to later on.

For example:

from sagemaker.pytorch import PyTorch
from sagemaker_ssh_helper.wrapper import SSHEstimatorWrapper  # <--NEW--

role = ...

estimator = PyTorch(entry_point='train.py',
                    source_dir='source_dir/training/',
                    dependencies=[SSHEstimatorWrapper.dependency_dir()],  # <--NEW--
                    role=role,
                    framework_version='1.9.1',
                    py_version='py38',
                    instance_count=1,
                    instance_type='ml.m5.xlarge')

ssh_wrapper = SSHEstimatorWrapper.create(estimator, connection_wait_time_seconds=600)  # <--NEW--

estimator.fit(wait=False)

instance_ids = ssh_wrapper.get_instance_ids()  # <--NEW--
print(f'To connect over SSM run: aws ssm start-session --target {instance_ids[0]}')  # <--NEW--

Note: connection_wait_time_seconds is the amount of time the SSH helper will wait inside SageMaker before it continues normal execution. It's useful for training jobs, when you want to connect before training starts. If you don't want to wait, set it to 0.

Note: If you use distributed training (i.e., instance_count > 1), SSH Helper will start by default only on the first 2 nodes (e.g., on algo-1 and algo-2). If you want to connect to SSH to other nodes, you can log in to either of these nodes, e.g., algo-1, and then SSH from this node to any other node of the training cluster, e.g., algo-4, without running SSH Helper on these nodes.

Alternatively, pass the additional parameter ssh_instance_count with the desired instance count to SSHEstimatorWrapper.create().

Note: if you a/ don't use script mode, b/ use basic Estimator class and c/ all code is already stored in your Docker container, check the code sample in the corresponding section of the FAQ.

Step 3: Modify your training script

Add into your train.py the following lines at the top:

import sagemaker_ssh_helper
sagemaker_ssh_helper.setup_and_start_ssh()

The setup_and_start_ssh() will start an SSM agent that will connect the training instance to AWS Systems Manager.

Step 4: Connect over SSM

Once you launched the job, you'll need to wait, a few minutes, for the SageMaker container to start and the SSM agent to start successfully. Then you'll need to have the ID of the managed instance. The instance id is prefixed by mi- and will appear in the job's CloudWatch log like this:

Successfully registered the instance with AWS SSM using Managed instance-id: mi-1234567890abcdef0

To fetch the instance IDs from the logs in an automated way, call the Python method of ssh_wrapper, as mentioned in the previous step:

estimator = ...
ssh_wrapper = ...
estimator.fit(wait=False)
instance_ids = ssh_wrapper.get_instance_ids()

This method accepts the optional parameter retry with the number of retry attempts (default is 60). It will retry to get instance IDs until they appear in the CloudWatch logs or number of attempts reached. The pause between attempts is fixed to 10 seconds, hence by default it will wait not more than 10 minutes.

With the instance id at hand, you will be able to connect to the training container using the command line or the AWS web console:

A. Connecting using command line:

On the local machine, make sure that you installed the latest AWS CLI v2 and the AWS Session Manager CLI plugin. Run the following command to perform the installation:

sm-local-configure

Note: If you start your training job from SageMaker Studio notebook, and execute the installation command from the image terminal, make sure that Python environment is activated, e.g., with conda activate base.

Run this command (replace the target value with the instance id for your SageMaker job). Example:

aws ssm start-session --target mi-1234567890abcdef0

B. Connecting using the AWS Web Console:

In AWS Web Console, navigate to Systems Manager > Fleet Manager.
Select the node, then Node actions > Start terminal session.

Once connected to the container, you would want to switch to the root user with sudo su - command.

Tip: Useful CLI commands

Here are some useful commands to run in a terminal session:

ps xfaww - Show running tree of processes
ps xfawwe - Show running tree of processes with environment variables
ls -l /opt/ml/input/data - Show input channels
ls -l /opt/ml/code - Show your training code
pip freeze | less - Show all Python packages installed
dpkg -l | less - Show all system packages installed

Tip: Generating a thread dump for stuck training jobs

In case your training job is stuck, it can be useful to observe what where its threads are waiting/busy. This can be done without connecting to a python debugger beforehand.

Having connected to the container as root, find the process id (pid) of the training process (assuming it's named train.py): pgrep --newest -f train.py
Install GNU debugger:
apt-get -y install gdb python3.9-dbg
Start the GNU debugger with python support:
gdb python
source /usr/share/gdb/auto-load/usr/bin/python3.9-dbg-gdb.py
Connect to the process (replace 361 with your pid):
attach 361
Show C low-level thread dump:
info threads
Show Python high-level thread dump:
py-bt
It might also be useful to observe what system calls the process is making: apt-get install strace
Trace the process (replace 361 with your pid):
sudo strace -p 361

Tip: Pipeline automation

If you're looking for the full automation of the pipeline with SSM and SSH, and not only with get_instance_ids() method, take a look at the automation question in the FAQ.

Connecting to SageMaker inference endpoints with SSM

Adding SageMaker SSH Helper to inference endpoint is similar to training with the following differences.

Wrap your model into SSHModelWrapper before calling deploy() and add SSH Helper to dependencies:

from sagemaker_ssh_helper.wrapper import SSHModelWrapper  # <--NEW--

estimator = ...
...
endpoint_name = ... 

model = estimator.create_model(entry_point='inference.py',
                               source_dir='source_dir/inference/',
                               dependencies=[SSHModelWrapper.dependency_dir()])  # <--NEW--

ssh_wrapper = SSHModelWrapper.create(model, connection_wait_time_seconds=0)  # <--NEW--

predictor = model.deploy(initial_instance_count=1,
                         instance_type='ml.m5.xlarge',
                         endpoint_name=endpoint_name)

predicted_value = predictor.predict(data=...)

Note: For the inference endpoint, which is always up and running, there's not too much value in setting connection_wait_time_seconds, so it's usually set to 0.

Similar to training jobs, you can fetch the instance ids for connecting to the endpoint with SSM with ssh_wrapper.get_instance_ids().

Add the following lines at the top of your inference.py script:

import os
import sys
sys.path.append(os.path.join(os.path.dirname(__file__), "lib"))

import sagemaker_ssh_helper
sagemaker_ssh_helper.setup_and_start_ssh()

Note: adding lib dir to Python path is required, because SageMaker inference is putting dependencies into the code/lib directory, while SageMaker training put libs directly to code.

Multi-model endpoints

For multi-model endpoints, the setup procedure is slightly different from regular endpoints:

from sagemaker.multidatamodel import MultiDataModel
from sagemaker_ssh_helper.wrapper import SSHModelWrapper, SSHMultiModelWrapper  # <--NEW--

model_data_prefix = "s3://DOC-EXAMPLE-BUCKET/mms/"
model_name = ...
endpoint_name = ...
estimator = ...
...

model = estimator.create_model(entry_point='inference.py',
                               source_dir='source_dir/inference/',
                               dependencies=[SSHModelWrapper.dependency_dir()])  # <--NEW--

mdm = MultiDataModel(
    name=model.name,
    model_data_prefix=model_data_prefix,
    model=model
)

ssh_wrapper = SSHMultiModelWrapper.create(mdm, connection_wait_time_seconds=0)  # <--NEW--

predictor = mdm.deploy(initial_instance_count=1,
                       instance_type='ml.m5.xlarge',
                       endpoint_name=endpoint_name)


mdm.add_model(model_data_source=model.repacked_model_data, model_data_path=model_name)

predicted_value = predictor.predict(data=..., target_model=model_name)

Important: Make sure that you're passing to add_model() the model ready for deployment with dependencies located at model.repacked_model_data, not the estimator.model_data that points to the trained model artifact. To obtain model suitable for inference, you might want to deploy first your model to a temporary single-node endpoint, so that SageMaker Python SDK takes care of repacking the model.

Also note that SageMaker SSH Helper will be lazy loaded together with your model upon the first prediction request. So you should try to connect to the multi-model endpoint only after calling predict().

The inference.py script is the same as for regular endpoints.

If you are using PyTorch containers, make sure you select the latest versions, e.g. 1.12, 1.11, 1.10 (1.10.2), 1.9 (1.9.1). This code might not work if you use PyTorch 1.8, 1.7 or 1.6.

Note: if you're packing your models manually and don't pass the model object to the MultiDataModel constructor, i.e., pass only the image_uri, see corresponding sample code in the FAQ.md.

Connecting to SageMaker batch transform jobs

For batch transform jobs, you need to use both SSHModelWrapper and SSHTransformerWrapper, as in the following example:

from sagemaker_ssh_helper.wrapper import SSHModelWrapper, SSHTransformerWrapper  # <--NEW--

sagemaker_session = ...
bucket = ...
estimator = ...
...

model = estimator.create_model(entry_point='inference.py',
                               source_dir='source_dir/inference/',
                               dependencies=[SSHModelWrapper.dependency_dir()])  # <--NEW--

transformer_input = sagemaker_session.upload_data(path='data/batch_transform/input',
                                                  bucket=bucket,
                                                  key_prefix='batch-transform/input')

transformer_output = f"s3://{bucket}/batch-transform/output"

ssh_model_wrapper = SSHModelWrapper.create(model, connection_wait_time_seconds=600)  # <--NEW--

transformer = model.transformer(instance_count=1,
                                instance_type="ml.m5.xlarge",
                                accept='text/csv',
                                strategy='SingleRecord',
                                assemble_with='Line',
                                output_path=transformer_output)

ssh_transformer_wrapper = SSHTransformerWrapper.create(transformer, ssh_model_wrapper)  # <--NEW--

transformer.transform(data=transformer_input,
                      content_type='text/csv',
                      split_type='Line',
                      join_source="Input",
                      wait=False)

The inference.py script is the same as for regular endpoints.

Connecting to SageMaker processing jobs

SageMaker SSH Helper supports both Script Processors and Framework processors and setup procedure is similar to training jobs and inference endpoints.

A. Framework processors

The code to set up a framework processor (e.g. PyTorch) is the following:

from sagemaker.pytorch import PyTorchProcessor
from sagemaker_ssh_helper.wrapper import SSHProcessorWrapper  # <--NEW--

role = ...

torch_processor = PyTorchProcessor(
    base_job_name='ssh-pytorch-processing',
    framework_version='1.9.1',
    py_version='py38',
    role=role,
    instance_count=1,
    instance_type="ml.m5.xlarge"
)

ssh_wrapper = SSHProcessorWrapper.create(torch_processor, connection_wait_time_seconds=600)  # <--NEW--

torch_processor.run(
    source_dir="source_dir/processing/",
    dependencies=[SSHProcessorWrapper.dependency_dir()],  # <--NEW--
    code="process_framework.py"
)

Also add the following lines at the top of process_framework.py:

import sagemaker_ssh_helper
sagemaker_ssh_helper.setup_and_start_ssh()

B. Script Processors

The code to set up a script processor (e.g. PySpark) is the following:

from sagemaker.spark import PySparkProcessor
from sagemaker_ssh_helper.wrapper import SSHProcessorWrapper  # <--NEW--

role = ...

spark_processor = PySparkProcessor(
    base_job_name='ssh-spark-processing',
    framework_version="3.0",
    role=role,
    instance_count=1,
    instance_type="ml.m5.xlarge"
)

ssh_wrapper = SSHProcessorWrapper.create(spark_processor, connection_wait_time_seconds=600)  # <--NEW--

spark_processor.run(
    submit_app="source_dir/processing/process.py",
    inputs=[ssh_wrapper.augmented_input()]  # <--NEW--
)

Also add the following lines at the top of process.py:

import sys
sys.path.append("/opt/ml/processing/input/")

import sagemaker_ssh_helper
sagemaker_ssh_helper.setup_and_start_ssh()

Remote debugging with PyCharm Debug Server over SSH

This procedure uses PyCharm's Professional feature: Remote debugging with the Python remote debug server configuration

On the local machine, make sure that you installed the latest AWS CLI v2 and the AWS Session Manager CLI plugin. Run the following command to perform the installation:

sm-local-configure

In PyCharm, go to the Run/Debug Configurations (Run -> Edit Configurations...), add a new Python Debug Server. Choose the fixed port, e. g. 12345.
Take the correct version of pydevd-pycharm package from the configuration window and install it either through requirements.txt or by calling pip from your source code.
Add commands to connect to the Debug Server to your code after the setup_and_start_ssh() (e.g., into a training script that you submit as an entry point for a training job):

import pydevd_pycharm
pydevd_pycharm.settrace('localhost', port=12345, stdoutToServer=True, stderrToServer=True, suspend=True)

Tip: Check the argument's description in the library source code.

Set extra breakpoints in your code with PyCharm, if needed
Start the Debug Server in PyCharm
Submit your code to SageMaker with SSH Helper as described in previous sections. Make sure you allow enough time for manually setting up the connection (do not set connection_wait_time_seconds to 0, recommended minimum value is 600, i.e. 10 minutes). Don't worry to set it to higher values, e.g. to 30 min, because you will be able to terminate the waiting loop once you connect.
Start the port forwarding script once SSH helper connects to SSM and starts waiting inside the training job:

sm-local-ssh-training connect <<training_job_name>>

It will reverse-forward the remote debugger port 12345 to your local machine's Debug Server port. The local port 11022 will be connected to the remote SSH server port, to allow you easily connect with SSH from command line.

Note: Before running this command make sure that AWS CLI is configured to access the account with aws s3 ls and your default region is set with aws configure command. Your Python's <path-to-venv>/bin/ directory should be in the $PATH, otherwise you will get a "command not found" error.

Tip: If you want to connect processing, batch transform jor or to an inference endpoint with SSH, use sm-local-ssh-processing, sm-local-ssh-transform or sm-local-ssh-inference scripts respectively.

While the sm-local-ssh-training script is running, you may connect with SSH to the specified local port (but it's not needed for PyCharm Debugger to work):

ssh -i ~/.ssh/sagemaker-ssh-gw -p 11022 root@localhost

Feel free to use the scripts as templates. Clone and customize them, if you want to change the ports.

Tip: If you log in to the node with SSH and don't see a sm-wait process, the training script has already started and failed to connect to the PyCharm Debug Server, so you need to increase the connection_wait_time_seconds, otherwise the debugger will miss your breakpoints.

Tip: If you don't want ssh command to complain about remote host keys, when you switch to a different node, consider adding this two options to the command: -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null.

Stop the waiting loop – connect to the instance and terminate the loop.

As already mentioned, make sure you've put enough timeout to allow the port forwarding script set up a tunnel before execution of your script continues.

You can use the following CLI command from your local machine to stop the waiting loop (the sm-wait remote process):

sm-local-ssh-training stop-waiting

Alternatively, if logged to the remote container already, run the pkill command from its shell (under root):

pkill sm-wait

After you stop the waiting loop, your code will continue running and will connect to your PyCharm Debug Server.

If everything is set up correctly, and you followed all the steps, PyCharm will stop at your breakpoint, highlight the line and wait for your input. Debug Server window will say “connected”. You can now press, for example, F8 to "Step Over" the code line or F7 to "Step Into" the code line.

Remote code execution with PyCharm / VSCode over SSH

Follow the steps from the section Remote debugging with PyCharm Debug Server: 1, 7, 8, 9.

Before terminating the waiting loop (step 9), make sure you configured and connected to SSH host and port localhost:11022 from your IDE as the remote Python interpreter. Provide ~/.ssh/sagemaker-ssh-gw as the private key.

Note, that after you finished the waiting loop, your training script will run only once, and you will be able to execute additional code only while your script is running. Once the script finishes, you will need to submit another training job and repeat the procedure again.

But there's a useful trick: submit a dummy script train_placeholder.py with the infinite loop, and while this loop will be running, you can run your real training script again and again with the remote interpreter. Setting max_run parameter of the estimator is highly recommended in this case.

The dummy script may look like this:

import time

import sagemaker_ssh_helper
sagemaker_ssh_helper.setup_and_start_ssh()

while True:
    time.sleep(10)

Make also sure that you're aware of SageMaker Managed Warm Pools feature, which is also helpful in such a scenario.

Pro Tip: Note that you can debug your code line by line in this scenario, too! See the tutorial in PyCharm documentation. Some users might prefer this option instead of using Debug Server as a simpler alternative.

Local IDE integration with SageMaker Studio over SSH for PyCharm / VSCode

Download Demo (.mov)

Follow the next steps for your local IDE integration with SageMaker Studio:

Copy SageMaker_SSH_IDE.ipynb into SageMaker Studio and run it.

Tip: Alternatively, instead of using SageMaker_SSH_IDE.ipynb, attach to a domain the KernelGateway lifecycle config script kernel-lc-config.sh (you may need to ask your administrator to do this). Once configured, from the Launcher choose the environment, puck up the lifecycle script and choose 'Open image terminal' (so, you don't even need to create a notebook).

On the local machine, install the library:

pip install sagemaker-ssh-helper

Make sure that you installed the latest AWS CLI v2 and the AWS Session Manager CLI plugin. Run the following command to perform the installation:

sm-local-configure

Start SSH tunnel and port forwarding from a terminal session as follows:

sm-local-ssh-ide connect <<kernel_gateway_app_name>>

The parameter <<kernel_gateway_app_name>> is either taken from SageMaker Studio when you run notebook SageMaker_SSH_IDE.ipynb, or from the image terminal as a hostname output, or you can find it in the list of running apps in AWS Console under Amazon SageMaker -> Control Panel -> User Details. It looks like this: sagemaker-data-science-ml-m5-large-1234567890abcdef0.

The local port 10022 will be connected to the remote SSH server port, to let you connect with SSH from IDE.
In addition, the local port 8889 will be connected to remote Jupyter notebook port, the port 5901 to the remote VNC server and optionally the remote port 443 will be connected to your local PyCharm license server address.

Note: Before running this command make sure that AWS CLI is configured to access the account with aws s3 ls and your default region is set with aws configure command. Your Python's <path-to-venv>/bin/ directory should be in the $PATH, otherwise you will get a "command not found" error.

Feel free to use the script as a template. Clone and customize it, if you want to change the ports and hosts.

Connect local PyCharm or VSCode with remote Python interpreter by using root@localhost:10022 as SSH parameters. Also provide ~/.ssh/sagemaker-ssh-gw as the private key.

Note: The SSH key is automatically generated on your local machine every time when you run sm-local-ssh-ide command from step 4.

Tip (PyCharm): When you configure Python interpreter in PyCharm, it's recommended to configure the path mapping ("Sync folders" deployment option) for you project to point into /root/project_name instead of default /tmp/pycharm_project_123. This is how you will be able to see your project in SageMaker Studio and PyCharm will automatically sync your local dir to the remote dir.

Tip (PyCharm): Also instead of creating a new venv, point the Python interpreter to the existing location. You can find this location by running a cell with import sys; sys.executable command in a SageMaker Studio notebook. You will get something like /opt/conda/bin/python.

Now with PyCharm or VSCode you can run and debug the code remotely inside the kernel gateway app.

You can also check from your local machine that connection is working by running the SSH command in command line:

ssh -i ~/.ssh/sagemaker-ssh-gw -p 10022 root@localhost

If you don't want this ssh command to complain about remote host keys, when you switch to a different node, consider adding this two options to the command: -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null.

You may now configure a remote Jupyter Server as http://127.0.0.1:8889/?token=<<your_token>>. You will find the full URL with remote token in the SageMaker_SSH_IDE.ipynb notebook in the output after running the cell with sm-ssh-ide start command.

Instructions for remote Jupyter notebooks in PyCharm
Instructions for remote Jupyter notebooks in VSCode (don't forget to switch kernel to remote after configuring the remote server).

You can also start the VNC session to vnc://localhost:5901 (e.g. on macOS with Screen Sharing app) and run IDE or any other GUI app on the remote desktop instead of your local machine.

If you want to switch to another kernel or instance, feel free to do so from SageMaker Studio UI and re-run SageMaker_SSH_IDE.ipynb.

Keep in mind that in this case the previous kernel will stop and SSM agent will stop, too. To allow multiple kernel and instances to be up and running with SageMaker SSH Helper and SSM agent, duplicate the notebook and give it a different name, e.g. SageMaker_SSH_IDE-PyTorch.ipynb. In this case you'll be able to keep two environments in parallel. To switch between them, you will only need to re-run sm-local-ssh-ide command on your local machine.

If you're using lifecycle configuration script, just start another image terminal with different environment settings from Launcher.

Don't forget to shut down SageMaker Studio resources, if you don't need them anymore, e.g., launched notebooks, terminals, apps and instances.

Troubleshooting IDE integration

Check that the managed instance in AWS Console in Systems Manager -> Fleet Manager section appears as "Online". Check that you're able to connect to the node from the Console by selecting Node actions -> Start terminal session.

If instance is "Offline", you might see this error message when calling an sm-local-ssh-ide command:

An error occurred (TargetNotConnected) when calling the StartSession operation: mi-1234567890abcdef0 is not connected.

or this one:

An error occurred (InvalidInstanceId) when calling the SendCommand operation: Instances [[mi-1234567890abcdef0]] not in a valid state for account 555555555555

Check SSM agent logs inside SageMaker Studio. From the image terminal run:

tail /var/log/amazon/ssm/*.log && date

Check that sshd process is started in SageMaker Studio notebook by running a command in the image terminal:

ps xfa | grep sshd

If it's not started, there might be some errors in the output of the notebook, and you might get this error on the local machine:

Connection closed by UNKNOWN port 65535

Check carefully the notebook output in SageMaker Studio to see if there are any installation or configuration problems that have to be fixed.

Sometimes you can see this error message on your local machine when trying to connect with SSM, even if you correctly completed all configuration steps, including the step to enable advanced tier:

An error occurred (BadRequest) when calling the StartSession operation: Enable advanced-instances tier to use Session Manager with your on-premises instances

In this case check that AWS Region in your AWS console is the same as on your local machine. To do this, run the following command locally:

aws configure list | grep region

It will provide you the locally configured region and will show where this setting is coming from, e.g., env variables or AWS config file:

    region                eu-west-1      config-file    ~/.aws/config

As the final effort, try to re-initialize the instance by restarting the notebook: Kernel -> Restart Kernel and Run All Cells.

Name		Name	Last commit message	Last commit date
Latest commit History 71 Commits
.idea		.idea
cdk		cdk
images		images
sagemaker_ssh_helper		sagemaker_ssh_helper
tests		tests
.gitignore		.gitignore
.gitlab-ci.yml		.gitlab-ci.yml
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
FAQ.md		FAQ.md
Flows.md		Flows.md
Flows_IDE.md		Flows_IDE.md
IAM_SSM_Setup.md		IAM_SSM_Setup.md
LICENSE		LICENSE
README.md		README.md
SageMaker_SSH_IDE.ipynb		SageMaker_SSH_IDE.ipynb
SageMaker_SSH_Notebook.ipynb		SageMaker_SSH_Notebook.ipynb
TODO		TODO
kernel-lc-config.sh		kernel-lc-config.sh
setup.py		setup.py

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

SageMaker SSH Helper

How it works

Getting started

Use Cases

Connecting to SageMaker training jobs with SSM

Step 1: Install the library

Step 2: Modify your start training job code

Step 3: Modify your training script

Step 4: Connect over SSM

Tip: Useful CLI commands

Tip: Generating a thread dump for stuck training jobs

Tip: Pipeline automation

Connecting to SageMaker inference endpoints with SSM

Multi-model endpoints

Connecting to SageMaker batch transform jobs

Connecting to SageMaker processing jobs

A. Framework processors

B. Script Processors

Remote debugging with PyCharm Debug Server over SSH

Remote code execution with PyCharm / VSCode over SSH

Local IDE integration with SageMaker Studio over SSH for PyCharm / VSCode

Troubleshooting IDE integration

About

Releases

Packages

Languages

License

nearmap/sagemaker-ssh-helper

Folders and files

Latest commit

History

Repository files navigation

SageMaker SSH Helper

How it works

Getting started

Use Cases

Connecting to SageMaker training jobs with SSM

Step 1: Install the library

Step 2: Modify your start training job code

Step 3: Modify your training script

Step 4: Connect over SSM

Tip: Useful CLI commands

Tip: Generating a thread dump for stuck training jobs

Tip: Pipeline automation

Connecting to SageMaker inference endpoints with SSM

Multi-model endpoints

Connecting to SageMaker batch transform jobs

Connecting to SageMaker processing jobs

A. Framework processors

B. Script Processors

Remote debugging with PyCharm Debug Server over SSH

Remote code execution with PyCharm / VSCode over SSH

Local IDE integration with SageMaker Studio over SSH for PyCharm / VSCode

Troubleshooting IDE integration

About

Resources

License

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages