Managed Jobs

Tip

This feature is great for scaling out: running a single job for long durations, or running many jobs in parallel.

SkyPilot supports managed jobs (sky jobs), which can automatically retry failures, recover from spot instance preemptions, and clean up when done.

To start a managed job, use sky jobs launch:

$ sky jobs launch -n myjob hello_sky.yaml

Task from YAML spec: hello_sky.yaml
Managed job 'myjob' will be launched on (estimated):
Considered resources (1 node):
------------------------------------------------------------------------------------------
 CLOUD   INSTANCE      vCPUs   Mem(GB)   ACCELERATORS   REGION/ZONE   COST ($)   CHOSEN
------------------------------------------------------------------------------------------
 AWS     m6i.2xlarge   8       32        -              us-east-1     0.38          ✔
------------------------------------------------------------------------------------------
Launching a managed job 'myjob'. Proceed? [Y/n]: Y
... <job is submitted and launched>
(setup pid=2383) Running setup.
(myjob, pid=2383) Hello, SkyPilot!
✓ Managed job finished: 1 (status: SUCCEEDED).

Managed Job ID: 1
📋 Useful Commands
├── To cancel the job:                sky jobs cancel 1
├── To stream job logs:               sky jobs logs 1
├── To stream controller logs:        sky jobs logs --controller 1
├── To view all managed jobs:         sky jobs queue
└── To view managed job dashboard:    sky jobs dashboard

The job is launched on a temporary SkyPilot cluster, managed end-to-end, and automatically cleaned up.

Managed jobs have several benefits:

1. Use spot instances: Jobs can run on auto-recovering spot instances. This saves significant costs (e.g., ~70% for GPU VMs) by making preemptible spot instances useful for long-running jobs.

2. Scale across regions and clouds: Easily run and manage thousands of jobs at once, using instances and GPUs across multiple regions/clouds.

3. Recover from failure: When a job fails, you can automatically retry it on a new cluster, eliminating flaky failures.

4. Managed pipelines: Run pipelines that contain multiple tasks. Useful for running a sequence of tasks that depend on each other, e.g., data processing, training a model, and then running inference on it.

Create a managed job

A managed job is created from a standard SkyPilot YAML. For example:

# bert_qa.yaml
name: bert-qa

resources:
  accelerators: V100:1
  use_spot: true  # Use spot instances to save cost.

envs:
  # Fill in your wandb key: copy from https://wandb.ai/authorize
  # Alternatively, you can use `--env WANDB_API_KEY=$WANDB_API_KEY`
  # to pass the key in the command line, during `sky jobs launch`.
  WANDB_API_KEY:

# Assume your working directory is under `~/transformers`.
# To get the code for this example, run:
# git clone https://github.com/huggingface/transformers.git ~/transformers -b v4.30.1
workdir: ~/transformers

setup: |
  pip install -e .
  cd examples/pytorch/question-answering/
  pip install -r requirements.txt torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
  pip install wandb

run: |
  cd examples/pytorch/question-answering/
  python run_qa.py \
    --model_name_or_path bert-base-uncased \
    --dataset_name squad \
    --do_train \
    --do_eval \
    --per_device_train_batch_size 12 \
    --learning_rate 3e-5 \
    --num_train_epochs 50 \
    --max_seq_length 384 \
    --doc_stride 128 \
    --report_to wandb \
    --output_dir /tmp/bert_qa/

Note

Workdir and file mounts with local files will be automatically uploaded to a cloud bucket. The bucket will be cleaned up after the job finishes.

To launch this YAML as a managed job, use sky jobs launch:

$ sky jobs launch -n bert-qa-job bert_qa.yaml

To see all flags, you can run sky jobs launch --help or see the CLI reference for more information.

SkyPilot will launch and start monitoring the job.

• Under the hood, SkyPilot spins up a temporary cluster for the job.

• If a spot preemption or any machine failure happens, SkyPilot will automatically search for resources across regions and clouds to re-launch the job.

• Resources are cleaned up as soon as the job is finished.

Tip

You can test your YAML on unmanaged sky launch , then do a production run as a managed job using sky jobs launch.

sky launch and sky jobs launch have a similar interface, but are useful in different scenarios.

sky launch (cluster jobs)	sky jobs launch (managed jobs)
Long-lived, manually managed cluster	Dedicated auto-managed cluster for each job
Spot preemptions must be manually recovered	Spot preemptions are auto-recovered
Number of parallel jobs limited by cluster resources	Easily manage hundreds or thousands of jobs at once
Good for interactive dev	Good for scaling out production jobs

Work with managed jobs

For a list of all commands and options, run sky jobs --help or read the CLI reference.

See a list of all managed jobs:

$ sky jobs queue

Fetching managed jobs...
Managed jobs:
ID NAME     RESOURCES           SUBMITTED   TOT. DURATION   JOB DURATION   #RECOVERIES  STATUS
2  roberta  1x [A100:8][Spot]   2 hrs ago   2h 47m 18s      2h 36m 18s     0            RUNNING
1  bert-qa  1x [V100:1][Spot]   4 hrs ago   4h 24m 26s      4h 17m 54s     0            RUNNING

Stream the logs of a running managed job:

$ sky jobs logs -n bert-qa  # by name
$ sky jobs logs 2           # by job ID

Cancel a managed job:

$ sky jobs cancel -n bert-qa  # by name
$ sky jobs cancel 2           # by job ID

Note

If any failure happens for a managed job, you can check sky jobs queue -a for the brief reason of the failure. For more details related to provisioning, check sky jobs logs --controller <job_id>.

Jobs dashboard

Use sky jobs dashboard to open a dashboard to see all jobs:

$ sky jobs dashboard

This automatically opens a browser tab to show the dashboard:

The UI shows the same information as the CLI sky jobs queue -a. The UI is especially useful when there are many in-progress jobs to monitor, which the terminal-based CLI may need more than one page to display.

Running on spot instances

Managed jobs can run on spot instances, and preemptions are auto-recovered by SkyPilot.

To run on spot instances, use sky jobs launch --use-spot, or specify use_spot: true in your SkyPilot YAML.

name: spot-job

resources:
  accelerators: A100:8
  use_spot: true

run: ...

Tip

Spot instances are cloud VMs that may be “preempted”. The cloud provider can forcibly shut down the underlying VM and remove your access to it, interrupting the job running on that instance.

In exchange, spot instances are significantly cheaper than normal instances that are not subject to preemption (so-called “on-demand” instances). Depending on the cloud and VM type, spot instances can be 70-90% cheaper.

SkyPilot automatically finds available spot instances across regions and clouds to maximize availability. Any spot preemptions are automatically handled by SkyPilot without user intervention.

Note

By default, a job will be restarted from scratch after each preemption recovery. To avoid redoing work after recovery, implement checkpointing and recovery. Your application code can checkpoint its progress periodically to a mounted cloud bucket. The program can then reload the latest checkpoint when restarted.

Here is an example of a training job failing over different regions across AWS and GCP.

Quick comparison between managed spot jobs vs. launching unmanaged spot clusters:

Command	Managed?	SSH-able?	Best for
sky jobs launch --use-spot	Yes, preemptions are auto-recovered	No	Scaling out long-running jobs (e.g., data processing, training, batch inference)
sky launch --use-spot	No, preemptions are not handled	Yes	Interactive dev on spot instances (especially for hardware with low preemption rates)

Either spot or on-demand/reserved

By default, on-demand instances will be used (not spot instances). To use spot instances, you must specify --use-spot on the command line or use_spot: true in your SkyPilot YAML.

However, you can also tell SkyPilot to use both spot instance and on-demand instances, depending on availability. In your SkyPilot YAML, use any_of to specify either spot or on-demand/reserved instances as candidate resources for a job. See documentation here for more details.

resources:
  accelerators: A100:8
  any_of:
    - use_spot: true
    - use_spot: false

In this example, SkyPilot will choose the cheapest resource to use, which almost certainly will be spot instances. If spot instances are not available, SkyPilot will fall back to launching on-demand/reserved instances.

Checkpointing and recovery

To recover quickly, a cloud bucket is typically needed to store the job’s states (e.g., model checkpoints). Any data on disk that is not stored inside a cloud bucket will be lost during the recovery process.

Below is an example of mounting a bucket to /checkpoint.

file_mounts:
  /checkpoint:
    name: # NOTE: Fill in your bucket name
    mode: MOUNT

The MOUNT mode in SkyPilot bucket mounting ensures the checkpoints outputted to /checkpoint are automatically synced to a persistent bucket.

To implement checkpointing in your application code:

1. Periodically save checkpoints containing the current program state during execution.

2. When starting/restarting, check if any checkpoints are present, and load the latest checkpoint.

Both features are included in most model training libraries, such as PyTorch, TensorFlow, and Hugging Face.

To see checkpointing in action, see the BERT end-to-end example below.

For other types of workloads, you can implement a similar mechanism as long as you can store the program state to/from disk.

Jobs restarts on user code failure

Preemptions or hardware failures will be auto-recovered, but by default, user code failures (non-zero exit codes) are not auto-recovered.

In some cases, you may want a job to automatically restart even if it fails in application code. For instance, if a training job crashes due to an NVIDIA driver issue or NCCL timeout, it should be recovered. To specify this, you can set max_restarts_on_errors in resources.job_recovery in the job YAML file.

resources:
  accelerators: A100:8
  job_recovery:
    # Restart the job up to 3 times on user code errors.
    max_restarts_on_errors: 3

This will restart the job, up to 3 times (for a total of 4 attempts), if your code has any non-zero exit code. Each restart runs on a newly provisioned temporary cluster.

When will my job be recovered?

Here’s how various kinds of failures will be handled by SkyPilot:

User code fails (setup or run commands have non-zero exit code):	If max_restarts_on_errors is set, restart up to that many times. If max_restarts_on_errors is not set, or we run out of restarts, set the job to FAILED or FAILED_SETUP.
Instances are preempted or underlying hardware fails:	Tear down the old temporary cluster and provision a new one in another region, then restart the job.
Can’t find available resources due to cloud quota or capacity restrictions:	Try other regions and other clouds indefinitely until resources are found.
Cloud config/auth issue or invalid job configuration:	Mark the job as FAILED_PRECHECKS and exit. Won’t be retried.

To see the logs of user code (setup or run commands), use sky jobs logs <job_id>. If there is a provisioning or recovery issue, you can see the provisioning logs by running sky jobs logs --controller <job_id>.

Tip

Under the hood, SkyPilot uses a “controller” to provision, monitor, and recover the underlying temporary clusters. See How it works: The jobs controller.

Examples

BERT end-to-end

We can take the SkyPilot YAML for BERT fine-tuning from above, and add checkpointing/recovery to get everything working end-to-end.

Note

You can find all the code for this example in the SkyPilot GitHub repository

In this example, we fine-tune a BERT model on a question-answering task with HuggingFace.

This example:

• has SkyPilot find a V100 instance on any cloud,

• uses spot instances to save cost, and

• uses checkpointing to recover preempted jobs quickly.

# bert_qa.yaml
name: bert-qa

resources:
  accelerators: V100:1
  use_spot: true  # Use spot instances to save cost.

file_mounts:
  /checkpoint:
    name: # NOTE: Fill in your bucket name
    mode: MOUNT

envs:
  # Fill in your wandb key: copy from https://wandb.ai/authorize
  # Alternatively, you can use `--env WANDB_API_KEY=$WANDB_API_KEY`
  # to pass the key in the command line, during `sky jobs launch`.
  WANDB_API_KEY:

# Assume your working directory is under `~/transformers`.
workdir: ~/transformers

setup: |
  pip install -e .
  cd examples/pytorch/question-answering/
  pip install -r requirements.txt torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
  pip install wandb

run: |
  cd examples/pytorch/question-answering/
  python run_qa.py \
    --model_name_or_path bert-base-uncased \
    --dataset_name squad \
    --do_train \
    --do_eval \
    --per_device_train_batch_size 12 \
    --learning_rate 3e-5 \
    --num_train_epochs 50 \
    --max_seq_length 384 \
    --doc_stride 128 \
    --report_to wandb \
    --output_dir /checkpoint/bert_qa/ \
    --run_name $SKYPILOT_TASK_ID \
    --save_total_limit 10 \
    --save_steps 1000

The highlighted lines add a bucket for checkpoints. As HuggingFace has built-in support for periodic checkpointing, we just need to pass the highlighted arguments to save checkpoints to the bucket. (See more on Huggingface API). To see another example of periodic checkpointing with PyTorch, check out our ResNet example.

We also set --run_name to $SKYPILOT_TASK_ID so that the logs for all recoveries of the same job will be saved to the same run in Weights & Biases.

Note

The environment variable $SKYPILOT_TASK_ID (example: “sky-managed-2022-10-06-05-17-09-750781_bert-qa_8-0”) can be used to identify the same job, i.e., it is kept identical across all recoveries of the job. It can be accessed in the task’s run commands or directly in the program itself (e.g., access via os.environ and pass to Weights & Biases for tracking purposes in your training script). It is made available to the task whenever it is invoked. See more about environment variables provided by SkyPilot.

With the highlighted changes, the managed job can now resume training after preemption! We can enjoy the benefits of cost savings from spot instances without worrying about preemption or losing progress.

$ sky jobs launch -n bert-qa bert_qa.yaml

Real-world examples

• Vicuna LLM chatbot: instructions, YAML

• Large-scale vector database ingestion, and the blog post about it

• BERT (shown above): YAML

• PyTorch DDP, ResNet: YAML

• PyTorch Lightning DDP, CIFAR-10: YAML

Scaling to many jobs

You can easily manage dozens, hundreds, or thousands of managed jobs at once. This is a great fit for batch jobs such as data processing, batch inference, or hyperparameter sweeps. To see an example launching many jobs in parallel, see Many Parallel Jobs.

To increase the maximum number of jobs that can run at once, see Best practices for scaling up the jobs controller.

Managed pipelines

A pipeline is a managed job that contains a sequence of tasks running one after another.

This is useful for running a sequence of tasks that depend on each other, e.g., training a model and then running inference on it. Different tasks can have different resource requirements to use appropriate per-task resources, which saves costs, while keeping the burden of managing the tasks off the user.

Note

In other words, a managed job is either a single task or a pipeline of tasks. All managed jobs are submitted by sky jobs launch.

To run a pipeline, specify the sequence of tasks in a YAML file. Here is an example:

name: pipeline

---

name: train

resources:
  accelerators: V100:8
  any_of:
    - use_spot: true
    - use_spot: false

file_mounts:
  /checkpoint:
    name: train-eval # NOTE: Fill in your bucket name
    mode: MOUNT

setup: |
  echo setup for training

run: |
  echo run for training
  echo save checkpoints to /checkpoint

---

name: eval

resources:
  accelerators: T4:1
  use_spot: false

file_mounts:
  /checkpoint:
    name: train-eval # NOTE: Fill in your bucket name
    mode: MOUNT

setup: |
  echo setup for eval

run: |
  echo load trained model from /checkpoint
  echo eval model on test set

The YAML above defines a pipeline with two tasks. The first name: pipeline names the pipeline. The first task has name train and the second task has name eval. The tasks are separated by a line with three dashes ---. Each task has its own resources, setup, and run sections. Tasks are executed sequentially. If a task fails, later tasks are skipped.

To pass data between the tasks, use a shared file mount. In this example, the train task writes its output to the /checkpoint file mount, which the eval task is then able to read from.

To submit the pipeline, the same command sky jobs launch is used. The pipeline will be automatically launched and monitored by SkyPilot. You can check the status of the pipeline with sky jobs queue or sky jobs dashboard.

$ sky jobs launch -n pipeline pipeline.yaml

$ sky jobs queue

Fetching managed job statuses...
Managed jobs
In progress jobs: 1 RECOVERING
ID  TASK  NAME      RESOURCES                    SUBMITTED    TOT. DURATION  JOB DURATION  #RECOVERIES  STATUS
8         pipeline  -                            50 mins ago  47m 45s        -             1            RECOVERING
 ↳  0     train     1x [V100:8][Spot|On-demand]  50 mins ago  47m 45s        -             1            RECOVERING
 ↳  1     eval      1x [T4:1]                    -            -              -             0            PENDING

Note

The $SKYPILOT_TASK_ID environment variable is also available in the run section of each task. It is unique for each task in the pipeline. For example, the $SKYPILOT_TASK_ID for the eval task above is: “sky-managed-2022-10-06-05-17-09-750781_pipeline_eval_8-1”.

Setting the job files bucket

For managed jobs, SkyPilot requires an intermediate bucket to store files used in the task, such as local file mounts, temporary files, and the workdir. If you do not configure a bucket, SkyPilot will automatically create a temporary bucket named skypilot-filemounts-{username}-{run_id} for each job launch. SkyPilot automatically deletes the bucket after the job completes.

Alternatively, you can pre-provision a bucket and use it as an intermediate for storing file by setting jobs.bucket in ~/.sky/config.yaml:

# ~/.sky/config.yaml
jobs:
  bucket: s3://my-bucket  # Supports s3://, gs://, https://<azure_storage_account>.blob.core.windows.net/<container>, r2://, cos://<region>/<bucket>

If you choose to specify a bucket, ensure that the bucket already exists and that you have the necessary permissions.

When using a pre-provisioned intermediate bucket with jobs.bucket, SkyPilot creates job-specific directories under the bucket root to store files. They are organized in the following structure:

# cloud bucket, s3://my-bucket/ for example
my-bucket/
├── job-15891b25/            # Job-specific directory
│   ├── local-file-mounts/   # Files from local file mounts
│   ├── tmp-files/           # Temporary files
│   └── workdir/             # Files from workdir
└── job-cae228be/            # Another job's directory
    ├── local-file-mounts/
    ├── tmp-files/
    └── workdir/

When using a custom bucket (jobs.bucket), the job-specific directories (e.g., job-15891b25/) created by SkyPilot are removed when the job completes.

Tip

Multiple users can share the same intermediate bucket. Each user’s jobs will have their own unique job-specific directories, ensuring that files are kept separate and organized.

How it works: The jobs controller

The jobs controller is a small on-demand CPU VM or pod running in the cloud that manages all jobs of a user. It is automatically launched when the first managed job is submitted, and it is autostopped after it has been idle for 10 minutes (i.e., after all managed jobs finish and no new managed job is submitted in that duration). Thus, no user action is needed to manage its lifecycle.

You can see the controller with sky status and refresh its status by using the -r/--refresh flag.

While the cost of the jobs controller is negligible (~$0.25/hour when running and less than $0.004/hour when stopped), you can still tear it down manually with sky down <job-controller-name>, where the <job-controller-name> can be found in the output of sky status.

Note

Tearing down the jobs controller loses all logs and status information for the finished managed jobs. It is only allowed when there are no in-progress managed jobs to ensure no resource leakage.

Customizing jobs controller resources

You may want to customize the resources of the jobs controller for several reasons:

1. Increasing the maximum number of jobs that can be run concurrently, which is based on the instance size of the controller. (Default: 90, see best practices)

2. Use a lower-cost controller (if you have a low number of concurrent managed jobs).

3. Enforcing the jobs controller to run on a specific location. (Default: cheapest location)

4. Changing the disk_size of the jobs controller to store more logs. (Default: 50GB)

To achieve the above, you can specify custom configs in ~/.sky/config.yaml with the following fields:

jobs:
  # NOTE: these settings only take effect for a new jobs controller, not if
  # you have an existing one.
  controller:
    resources:
      # All configs below are optional.
      # Specify the location of the jobs controller.
      cloud: gcp
      region: us-central1
      # Bump cpus to allow more managed jobs to be launched concurrently. (Default: 4+)
      cpus: 8+
      # Bump memory to allow more managed jobs to be running at once.
      # By default, it scales with CPU (8x).
      memory: 64+
      # Specify the disk_size in GB of the jobs controller.
      disk_size: 100

The resources field has the same spec as a normal SkyPilot job; see here.

Note

These settings will not take effect if you have an existing controller (either stopped or live). For them to take effect, tear down the existing controller first, which requires all in-progress jobs to finish or be canceled.

To see your current jobs controller, use sky status.

$ sky status --refresh

Clusters
NAME                          LAUNCHED     RESOURCES                          STATUS   AUTOSTOP  COMMAND
my-cluster-1                  1 week ago   1x AWS(m6i.4xlarge)                STOPPED  -         sky launch --cpus 16 --cloud...
my-other-cluster              1 week ago   1x GCP(n2-standard-16)             STOPPED  -         sky launch --cloud gcp --...
sky-jobs-controller-919df126  1 day ago    1x AWS(r6i.xlarge, disk_size=50)   STOPPED  10m       sky jobs launch --cpus 2 ...

Managed jobs
No in-progress managed jobs.

Services
No live services.

In this example, you can see the jobs controller (sky-jobs-controller-919df126) is an r6i.xlarge on AWS, which is the default size.

To tear down the current controller, so that new resource config is picked up, use sky down.

$ sky down sky-jobs-controller-919df126

WARNING: Tearing down the managed jobs controller. Please be aware of the following:
 * All logs and status information of the managed jobs (output of `sky jobs queue`) will be lost.
 * No in-progress managed jobs found. It should be safe to terminate (see caveats above).
To proceed, please type 'delete': delete
Terminating cluster sky-jobs-controller-919df126...done.
Terminating 1 cluster ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00

The next time you use sky jobs launch, a new controller will be created with the updated resources.

Best practices for scaling up the jobs controller

Tip

For managed jobs, it’s highly recommended to use service accounts for cloud authentication. This is so that the jobs controller credentials do not expire. This is particularly important in large production runs to avoid leaking resources.

The number of active jobs that the controller supports is based on the controller size. There are two limits that apply:

• Actively launching job count: maxes out at 4 * vCPU count. A job counts towards this limit when it is first starting, launching instances, or recovering.

  • The default controller size has 4 CPUs, meaning 16 jobs can be actively launching at once.

• Running job count: maxes out at memory / 350MiB, up to a max of 2000 jobs.

  • The default controller size has 32GiB of memory, meaning around 90 jobs can be running in parallel.

The default size is appropriate for most moderate use cases, but if you need to run hundreds or thousands of jobs at once, you should increase the controller size.

For maximum parallelism, the following configuration is recommended:

jobs:
  controller:
    resources:
      # In our testing, aws > gcp > azure
      cloud: aws
      cpus: 128
      # Azure does not have 128+ CPU instances, so use 96 instead
      # cpus: 96
      memory: 600+
      disk_size: 500

Note

Remember to tear down your controller to apply these changes, as described above.

With this configuration, you’ll get the following performance:

Cloud	Instance type	Launching jobs	Running jobs
AWS	r6i.32xlarge	512 launches at once	2000 running at once
GCP	n2-highmem-128	512 launches at once	2000 running at once
Azure	Standard_E96s_v5	384 launches at once	1930 running at once