Server connections interrupted

[patrik-turi]

Describe the bug

After sleeping/waking up the mac, I'm greeted with "Server connections interrupted" window, it doesn't matter if I click "Ignore" or "Disconnect".

Docker commands don't respond anymore just hang.
If I restart OrbStack, it's stuck at the "Starting..." spinner.

In the logs we found this error:

:evergreen_tree: vmgr | time="05-15 14:20:46" level=error msg="failed to send fsnotify events" error="write tcp 0.250.250.1:28550: endpoint is closed for send"

So far only restarting the mac has fixed the issue (until it gets stuck again after sleep).

To Reproduce

It seems to be related to sleep/wake, I had a single container running (wifi also changed).

Expected behavior

No response

Diagnostic report (REQUIRED)

OrbStack info:
Version: 1.10.3
Commit: 2b5dd5f580d80a3d2494b7b40dde2ef46813cfc5 (v1.10.3)

System info:
macOS: 15.4.1 (24E263)
CPU: arm64, 10 cores
CPU model: Apple M1 Max
Model: MacBookPro18,2
Memory: 32 GiB

Full report: https://orbstack.dev/_admin/diag/orbstack-diagreport_2025-05-15T12-20-06.877743Z.zip

Screenshots and additional context (optional)

No response

[shenxc087932]

I've also encountered this problem several times

[Taytay]

I am getting this problem every time after waking my Macbook. I can get it back by restarting Orbstack twice. (Quitting Orbstsack, starting it again, quitting it, and restarting it for a final time.)
In my case, I am on MacOS Tahoe Beta 26.0

Similar issue.

After stopped and started the orbstack, i cannot access my home network (192.168..) , using guacamole to rdp my windows , but the internet is work. If I ping the windows host, it shows 0.250.250.1

Even restarted orbstack multiple times , still not fix, only reboot the mac works.

My workaround is avoid to stop orbstack, but only stop all containers :(

[rogdex24]

+1 , Same Problem on MacOs 26.1
@Taytay restarting twice does work though, a little bit better solution that restarting the whole machine

[rogdex24]

Surprisingly much older issue that there have been many duplicates - #1342

[majordoobie]

I'm experiencing what appears to be a more severe variant of this issue. Instead of just the OrbStack host-side network breaking, I'm seeing full Linux VM kernel deadlock with virtio network driver failure.

My Setup

• Hardware: M2 Mac Mini (8 cores)
• macOS: 15.4 (24E248)
• OrbStack: Latest version
• Workload: Docker Compose stack with Gluetun VPN container using network_mode: service:gluetun to route 3+ other containers (qbittorrent, prowlarr, flaresolverr) through the VPN tunnel

Symptoms

• orb status shows "Starting" indefinitely
• All docker commands hang/timeout
• OrbStack Helper process consumes ~200% CPU
• System reports "1 stuck process"
• Requires force kill of OrbStack processes to recover

Key Log Evidence

1. Virtio network transmit queue timeout (VM kernel level):

👾 kernel | [11435.368086] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 650045 ms
👾 kernel | [11435.368115] virtio_net virtio8 eth0: TX timeout on queue: 0, sq: output.0, vq: 0x1, name: output.0, 650045000 usecs ago

2. RCU CPU stall (kernel deadlock):

👾 kernel | [11435.451086] rcu: INFO: rcu_sched self-detected stall on CPU
👾 kernel | [11435.451107] rcu: 	7-....: (629641 ticks this GP) idle=15c4/1/0x4000000000000000 softirq=24568/24568 fqs=143270
👾 kernel | [11435.451126] rcu: 	(t=670154 jiffies g=65213 q=12292 ncpus=8)

3. OrbStack internal network failure:

🌲 vmgr | time="01-13 11:04:10" level=error msg="host-unix forward: dial failed" addr="{1 0.250.250.2 2375 }" error="context deadline exceeded"
🌲 vmgr | time="01-13 11:04:28" level=error msg="host-unix forward: dial failed" addr="{1 0.250.250.2 8000 }" error="context deadline exceeded"

4. OrbStack detects VM hang:

🌲 vmgr | time="01-13 11:05:01" level=error msg="health check failed" error="Post \"http://vcontrol/disk/report_stats\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"
🌲 vmgr | time="01-13 11:05:01" level=warning msg="sampling stacks due to VM hang"

5. Repeated TCP forward failures:

🌲 vmgr | time="01-13 11:03:53" level=error msg="host-tcp forward: dial failed" addr="{1 fd07:b51a:cc66:f0::2 36625 }" error="context deadline exceeded"

Analysis

The pattern shows:

1. Virtio network transmit queue gets stuck (650+ seconds timeout)
2. This causes CPU 7 to deadlock in the VM kernel (RCU stall)
3. OrbStack's internal network on 0.250.250.x fails
4. Docker API becomes unreachable
5. Entire VM enters hung state

Unique Aspects

Unlike other reports where this seems to be triggered by sleep/wake cycles, mine occurs during normal heavy network operation with:

• VPN tunnel (Gluetun with OpenVPN)
• Multiple containers sharing a single network namespace via network_mode: service:gluetun
• Constant HTTP requests through the shared network and tons of data being downloaded.

This suggests the virtio network driver may have issues handling high-throughput, complex network scenarios beyond just state management during sleep/wake transitions.

Trigger Frequency

This happens predictably within 3-4 hours of running my container stack, which makes me believe it's related to cumulative network stress rather than random timing.

Full vmgr.log excerpt

Click to expand full log sequence showing progression from network timeout to kernel deadlock

[2m2026-01-13T16:03:53.407481Z[0m [32m INFO[0m [2mvmm::macos::vstate[0m[2m:[0m vCPU 2 state:
------ vCPU 2 ------
PC: 0xffffb143d9ecc94c      LR: 0xffffb143d9ecc964  FP: 0xffff50aa403d7df0
SP_EL1: 0xffff50aa403d7df0  TPIDR_EL1: 0xffff9f68e37a6000  VBAR_EL1: 0xffffb143d8c10800
ESR_EL1: 0x0000000056000000  FAR_EL1: 0x0000ffff7e21e000
PSTATE(el): EL1h

Stack:
  cpu_do_idle+4
  arch_cpu_idle+11
  default_idle_call+71
  cpu_startup_entry+247
  secondary_start_kernel+235
  __secondary_switched+115

🌲 vmgr | time="01-13 11:03:53" level=error msg="host-tcp forward: dial failed" addr="{1 fd07:b51a:cc66:f0::2 36625 }" error="context deadline exceeded"
🌲 vmgr | time="01-13 11:03:54" level=warning msg="stacks sampled"
👾 kernel | [11259.231442] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 473917 ms
👾 kernel | [11259.231466] virtio_net virtio8 eth0: TX timeout on queue: 0, sq: output.0, vq: 0x1, name: output.0, 473917000 usecs ago
👾 kernel | [11264.351693] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 479037 ms
👾 kernel | [11264.351722] virtio_net virtio8 eth0: TX timeout on queue: 0, sq: output.0, vq: 0x1, name: output.0, 479037000 usecs ago
🌲 vmgr | time="01-13 11:04:07" level=error msg="host-tcp forward: dial failed" addr="{1 fd07:b51a:cc66:f0::2 36625 }" error="context deadline exceeded"
👾 kernel | [11270.495995] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 485181 ms
👾 kernel | [11270.496033] virtio_net virtio8 eth0: TX timeout on queue: 0, sq: output.0, vq: 0x1, name: output.0, 485181000 usecs ago
🌲 vmgr | time="01-13 11:04:10" level=error msg="host-unix forward: dial failed" addr="{1 0.250.250.2 2375 }" error="context deadline exceeded"
👾 kernel | [11276.128270] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 490813 ms
🌲 vmgr | time="01-13 11:04:28" level=error msg="host-unix forward: dial failed" addr="{1 0.250.250.2 8000 }" error="context deadline exceeded"
👾 kernel | [11435.368086] virtio_net virtio8 eth0: NETDEV WATCHDOG: CPU: 7: transmit queue 0 timed out 650045 ms
👾 kernel | [11435.451086] rcu: INFO: rcu_sched self-detected stall on CPU
👾 kernel | [11435.451107] rcu: 	7-....: (629641 ticks this GP) idle=15c4/1/0x4000000000000000 softirq=24568/24568 fqs=143270
🌲 vmgr | time="01-13 11:05:01" level=error msg="health check failed" error="Post \"http://vcontrol/disk/report_stats\": context deadline exceeded (Client.Timeout exceeded while awaiting headers)"
🌲 vmgr | time="01-13 11:05:01" level=warning msg="sampling stacks due to VM hang"

Hope this additional data point helps narrow down the root cause. Happy to provide any additional diagnostics if needed.

[nikspyratos]

In my case, seems my VPN is doing something weird with localhost routing and didn't properly release that when it got turned off either. Rebooting solved it.

macOS Tahoe 26.1

[niieani]

Also seeing this happen

[zinzocom]

I've been hitting this as well; it happens when running write-heavy workloads like Grafana, Prometheus, or similar logging/monitoring stacks locally on macOS.

Looking at the crash report, the root cause is macOS disk write throttling:

• Event: disk writes
• Writes: 2,147.60 MB of file-backed memory dirtied over ~21,064 seconds (~5.8 hours), averaging ~101.95 KB/s
• macOS limit: ~24.86 KB/s over a rolling 86,400-second (24-hour) window, which works out to a ~2.1 GB budget
• Result: OrbStack Helper exceeded the 24-hour write budget in roughly 5.8 hours (writing at ~4x the allowed rate), so macOS killed the process

The heaviest stack trace points to pwritev, which lines up with the container workloads; Prometheus TSDB compaction, Postgres WAL/checkpoints, and Grafana are all continuously writing to disk.

Workarounds that helped on my end:

1. Reduce Prometheus write pressure: lower the scrape frequency and shorten retention. TSDB compaction is one of the biggest contributors.
2. Tune Postgres checkpointing: increase checkpoint_completion_target and checkpoint_timeout to spread writes over longer intervals instead of short bursts.
3. Move the OrbStack data disk to an external volume: macOS tracks write budgets per filesystem, so offloading to a separate drive sidesteps the limit entirely.
4. Disable macOS disk write throttling for OrbStack: this is the nuclear option but it works if the above aren't enough.

Hope this helps narrow it down.