windows-2022: docker not available when runner starts

[sorindumitru]

Description

We've started seeing very frequent errors from our actions running on windows. docker doesn't seem to be available when we want to use it:

failed to connect to the docker API at npipe:////./pipe/docker_engine; check if the path is correct and if the daemon is running: open //./pipe/docker_engine: The system cannot find the file specified.

It used to work ok early last, e.g. on the 17th, but now it fails most of the time, but not always.

Platforms affected

• Azure DevOps
• GitHub Actions - Standard Runners
• GitHub Actions - Larger Runners

Runner images affected

• Ubuntu 22.04
• Ubuntu 24.04
• Ubuntu Slim
• macOS 14
• macOS 14 Arm64
• macOS 15
• macOS 15 Arm64
• macOS 26
• macOS 26 Arm64
• Windows Server 2022
• Windows Server 2025
• Windows Server 2025 with Visual Studio 2026

Image version and build link

https://github.com/spiffe/spire/actions/runs/22362890751/job/64722353668?pr=6677

Is it regression?

yes

Expected behavior

Docker should be available

Actual behavior

Docker is complaining that it can't find the named pipe to connect to docker engine.

Repro steps

1. In a runner attempt to use docker.

[sorindumitru]

Looks like if we wait for 60 seconds, it still doesn't become ready, the status of the docker service seems to be stopped:

Status  Name   DisplayName
------  ----   -----------
Stopped docker Docker Engine

If we start it using:

Start-Service $_.Name -ErrorAction SilentlyContinue

it eventually starts working.

[v-SergeiRadich]

Hi!

We will check it and return with details soon.

[v-SergeiRadich]

👋

We have tried to reproduce the issue in test workflow, but it happened only 1 time out of 3x10 runs for windows-2022. It looks like random bad host issue which can be handled by job rerun.

Please reply if you need any more assistance.

[sorindumitru]

Hi @v-SergeiRadich, thanks for looking into it. This happens very frequently for us. It's probably much closer to 50/50. Here's a couple more instances (and I can provide a lot more, but I need to dig through previous runs):
https://github.com/spiffe/spire/actions/runs/22362890751/job/64722353668
https://github.com/spiffe/spire/actions/runs/22320310813/job/64577658266

This is actively preventing us from merging changes since we run into this issue very frequently.

Wouldn't it be safe to say that something should be making sure that Docker is up when the runner starts the job?

[v-SergeiRadich]

We'll proceed with additional investigation and reply once we'll have more information.

[jonny64bit]

This is happening 50% of the time for us. Only extra info to add is we are getting this on the large windows runners.

Here is a log with a few bits redacted.

job-logs.txt

[sorindumitru]

We're currently starting docker if it is not started and that seems to make it work reliably. But that shouldn't be required, it should just be available.