hashicorp/vagrant

Crashes Mac

davejlong opened this issue · 7 comments

I run a multi-vm environment with 3 base boxes running different sets of software. The boxes were created with a previous release of Vagrant when it was a ruby gem. At times, when I have different combinations of the boxes running, the Mac will completely crash. I will post the report provided next time it crashes (usually a couple times a day).

This is a known VirtualBox issue. VirtualBox kernel panics too frequently. To address this, I will be working on other hypervisor support in the very near future (such as VMWare Fusion).

I've been getting kernel panics a lot recently while shutting down VMs (same deal, multiple VMs with host only networking, talking to one another).

I think it's this issue: https://www.virtualbox.org/ticket/9359

We use WiFi and an Airport in the office for our network. According to the above linked issue this problem is related to that. I'm trying out going hardwired with ethernet and wifi disabled.. I'll report back about wether or not this helps the problem.

I've seen this sort of instability on OSX, and it makes no odds whether I'm using wifi or wired networking. I've fed a number of crash dumps and other details to one of the VirtualBox developers after a bit of email communication, but he's been unable to solve the problem thus far :(

As per https://www.virtualbox.org/ticket/9897, I've found that downgrading to 4.0.16 seems to solve this problem. I don't know if there's anything in 4.0 that's incompatible with Vagrant, maybe @mitchellh can comment?

VirtualBox just all around crashes quite a bit, unfortunately. "Quite a bit" seems to vary from person to person. For example, at my work, somehow, no one ever kernel panics. I see a kernel panic about once a week but I'm a pretty heavy user of Vagrant ;)

They claim to have fixed this bug.

VMM: fixed a potential host crash triggered by shutting down a VM when another VM was running (only affected 32-bit hosts and 64-bit OS X hosts, 4.1 regression, bug #9897)

From VirtualBox changelog version 4.1.22
This was still happening to me on a box shutdown with at least one other box running, wired connection
Can't wait to see if this fixes it!

I have a crash test script which starts one VM, and cycles another one on and off repeatedly. Previously, this has been able to panic the machine within about 20 or 30 iterations. I've managed almost 800 so far, with 4.1.22, which is extremely promising.