Describe the bug

OrbStack VM manager is repeatedly killed under Docker dev workload

Summary

OrbStack repeatedly stops while running a local Docker-based development stack. The user-visible symptom is that Docker becomes unreachable and all forwarded container ports disappear. The OrbStack GUI reports server connections interrupted. The application then sees database/Redis connection failures even though the containers had previously started successfully.

This appears to be an OrbStack VM/port-forwarding/file-sharing crash or host-side kill, not an application-level Postgres or Redis failure.

Environment

• macOS on Apple Silicon
• Machine memory: 32 GB
• OrbStack: 2.2.0
• Docker engine reported by OrbStack before failure: 29.4.0
• OrbStack settings tried:
  • cpu: 6
  • memory_mib: 24576 originally, then 12288, then 8192
  • k8s.enable: false
  • power.pause_in_sleep: false

Workload

Local development stack with Docker Compose:

• Veritly infra:
  • Postgres on host port 5432
  • MinIO on host ports 9000, 9001
• Activepieces infra:
  • Postgres on host port 5433
  • Redis on host port 6380
• Host-side Node/Bun/Vite processes:
  • Bun backend
  • Vite frontend
  • Activepieces API, worker, engine
  • TypeScript/Turbo piece builds

The containers can start and report healthy briefly. For example:

docker-postgres-1 Up ... (healthy) 0.0.0.0:5432->5432/tcp
automation-activepieces-db-1 Up ... (healthy) 0.0.0.0:5433->5432/tcp
automation-activepieces-redis-1 Up ... (healthy) 0.0.0.0:6380->6379/tcp
redis-cli ping -> PONG

Then OrbStack stops and Docker becomes unreachable.

Observed Failure

Application logs after OrbStack disappears:

ioredis Unhandled error event: Error: connect ECONNREFUSED 127.0.0.1:6380
pg error: Connection terminated unexpectedly
nfs server OrbStack:/OrbStack: not responding
Cannot connect to the Docker daemon at unix:///Users/Apple/.orbstack/run/docker.sock.

OrbStack status after failure:

Stopped

macOS / OrbStack logs show:

Daemon exited: killed (SIGKILL)
Host exit diagnostics:
- reason: killed (SIGKILL)
- raw wait status: 9 (0x00000009)
- raw flags: 0x86000000
- exit detail: none

OrbStack also saved:

/Users/Apple/.orbstack/log/unified-kill.log

The vmgr.log around startup shows NFS and Docker port forwarding activity:

service is ready service=nfs
Mounting NFS...
NFS mounted
add forward container=docker spec="tcp:0.0.0.0:5432"
add forward container=docker spec="tcp:0.0.0.0:9000"
add forward container=docker spec="tcp:0.0.0.0:9001"
add forward container=docker spec="tcp:0.0.0.0:5433"
add forward container=docker spec="tcp:0.0.0.0:6380"

Then after failure, forwarded ports vanish and Docker is gone:

remove forward key="tcp:0.0.0.0:5433"
remove forward key="tcp:0.0.0.0:6380"
Cannot connect to the Docker daemon at unix:///Users/Apple/.orbstack/run/docker.sock.

There were also suspicious port-forward binding errors for an internal address restored from an unrelated container/network:

failed to add forward container=docker error="host: listen tcp4 0.250.250.65:<port>: bind: can't assign requested address"
failed to add forward container=docker error="host: listen tcp [fd07:b51a:cc66:f0::41]:<port>: bind: can't assign requested address"

Removing that unrelated container/network did not stop the SIGKILL failure.

Useful Files

• /Users/Apple/.orbstack/log/gui.log
• /Users/Apple/.orbstack/log/vmgr.log
• /Users/Apple/.orbstack/log/unified-kill.log
• /Users/Apple/.orbstack/vmconfig.json

gui.log
vmgr.log
vmgr.1.log
unified-kill.log

vmgr.log

To Reproduce

Reproduction Shape

1. Start OrbStack.
2. Ensure Docker context is orbstack.
3. Start Docker Compose infra for Postgres, MinIO, Postgres, Redis.
4. Start host-side Node/Bun dev processes that connect to those containers.
5. Run Activepieces setup/dev startup, including Turbo/TypeScript piece builds and API/worker/engine boot.
6. Within roughly 10-30 seconds, OrbStack stops. Docker daemon socket disappears.

Expected behavior

What we ruled out

• Not a simple host port conflict:
  • Postgres and Redis containers bind successfully before the failure.
  • Redis responds to PONG before OrbStack stops.
• Not Kubernetes:
  • Reproduced after k8s.enable: false and restart.
• Not just excessive VM memory:
  • Reproduced with memory_mib lowered from 24 GB to 12 GB and then 8 GB.

Hypothesis

The OrbStack VM manager or its host-side Docker/NFS/port-forwarding layer is being killed under a bursty local development workload involving:

• several published container ports,
• NFS/file sharing active,
• Docker networks being created/removed,
• simultaneous host-side TypeScript/Turbo builds and Node service startup.

The primary actionable signal is the host-side vmgr SIGKILL:

vmgrExit(reason: killed (SIGKILL), raw wait status 9)

The app-level Redis/Postgres errors are downstream symptoms caused by OrbStack stopping.

Diagnostic report (REQUIRED)

OrbStack info:
Version: 2.2.0
Commit: 0cd235b96a29195592ca86fbababff23d67abfd6 (v2.2.0)

System info:
macOS: 15.7.5 (24G617)
CPU: arm64, 10 cores
CPU model: Apple M1 Max
Model: MacBookPro18,2
Memory: 32 GiB

Full report: https://orbstack.dev/_admin/diag/orbstack-diagreport_2026-06-04T14-38-02.752210Z.zip

Screenshots and additional context (optional)

No response