[Bugfix] Fix some cases of The client socket has timed out after 600s while trying to connect to

[guoyuhong]

vLLM works fine in most of my environments. However, there are always some envs that have following error:
The client socket has timed out after 600s while trying to connect to

In this case, I need to set the environment variable export VLLM_HOST_IP="127.0.0.1". However, for a new user, he/she may not realize what happens.
After digging into this problem, I found that this is caused by the automatically set environment variable HOST_IP.

In many cases, we use vLLM in a container. The container may use the Bridge Mode for the network, which means the container has different IP from the host machine. And sometimes, the container is started with the environment variable HOST_IP set to the host machine's IP. In current logic of get_ip() function and env.VLLM_HOST_IP, if HOST_IP is set but VLLM_HOST_IP is not set, get_ip() will return HOST_IP without trying to get a real IP.

The solution is to try the best to find the local IP first. The env HOST_IP is used only when we cannot find the local IP from code and VLLM_HOST_IP is not set.

May FIX #5779 #6650 #7136

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[guoyuhong]

This PR is ready for review.

[guoyuhong]

@DarkLight1337 could you take a look at this PR?

[youkaichao]

the container is started with the environment variable HOST_IP set to the host machine's IP

any reference for this? is it common?

[guoyuhong]

the container is started with the environment variable HOST_IP set to the host machine's IP

any reference for this? is it common?

@youkaichao This is not a standard but it is used in many systems, since it is a common name. e.g.

1. https://sap.github.io/cf-service-operator/docs/development/local/
2. https://youtrack.jetbrains.com/issue/WI-42003/Environment-Variables-for-docker-compose-are-not-usable-or-not-taken-into-account
3. https://docs.openstack.org/devstack/latest/configuration.html
4. [stable/ambassador] Add HOST_IP environment variable to ambassador container helm/charts#13555

[youkaichao]

HOST_IP is kept mainly for historical reasons. I think it is time to remove it.

[youkaichao]

@guoyuhong please check #10644

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @guoyuhong.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[guoyuhong]

Close this PR since #10644 is merged.