Skip to content

update checkpointing docs #282

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yaseenisolated merged 3 commits into from
Jun 17, 2026
Merged

update checkpointing docs #282

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
724531a
update checkpointing docs
yaseenisolated Jun 11, 2026
79120af
remove trailing slash
yaseenisolated Jun 11, 2026
e1211bc
fix except
yaseenisolated Jun 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 22 additions & 7 deletions other-topics/checkpointing.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,8 +16,8 @@ Cerebrium has native checkpointing and restore functionality built in to the pla
Checkpointing is available on our v2 runtime environment. Add the following to your `cerebrium.toml` to upgrade.

```
[cerebrium.runtime]
container_runtime = "v2"
[cerebrium.experimental]
checkpointing = true
```

To create a checkpoint your application has to send a trigger to our runtime after it has performed its initialization and is ready. When this trigger is received, the runtime verifies if a new checkpoint is required. To save resources, the system will not create a new checkpoint if:
Expand All @@ -38,17 +38,28 @@ A checkpoint is tightly coupled to a single deployment. To disable restoring fro
```python
from vllm import AsyncLLMEngine
from vllm.engine.arg_utils import AsyncEngineArgs
import http
import urllib

# Init vLLM engine
engine_args = AsyncEngineArgs(
model="Qwen/Qwen2.5-0.5B-Instruct",
async_scheduling=False
async_scheduling=False,
sleep_mode=True
)
AsyncLLMEngine.from_engine_args(engine_args)
engine = AsyncLLMEngine.from_engine_args(engine_args)

# Drop KV cache for reduced GPU memory footprint.
engine.sleep(level=1)
# Trigger checkpoint
urllib.request.urlopen("http://169.254.169.253:8234/checkpoint/", method="POST")
# Wait for it to complete
urllib.request.urlopen("http://169.254.169.253:8234/checkpoint/wait")
try:
urllib.request.urlopen("http://169.254.169.253:8234/checkpoint", method="POST")
except http.client.RemoteDisconnect:
# TCP connections disconnect on restore and throw remote
pass

# Restore KV cache
engine.wake_up()
```

## Limitations
Expand All @@ -70,3 +81,7 @@ urllib.request.urlopen("http://169.254.169.253:8234/checkpoint/wait")
vLLM checkpointing support is not complete but still possible. See https://github.com/vllm-project/vllm/issues/34303 and other issues.

If you are getting an EngineCoreDead exception add `async_scheduling=False` to your AsyncEngineArgs and it should succeed.

The larger the size of the memory checkpoint the slower the restore is. We can reduce the size of the snapshot substantially and improve startup times by dropping the KV Cache before checkpoint and recreating it after restore. vLLM has functionality that does this built in as part of [vLLM Sleep Mode](https://docs.vllm.ai/en/latest/features/sleep_mode/).

@jonoirwinrsa jonoirwinrsa Jun 11, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

are users able to see the size of the snapshot? Do we log it?


You
Loading