Dainis Berzins
Dabbler
- Joined
- Aug 29, 2014
- Messages
- 29
I'm having issues with VirtualBox when booting Linux distros on newly created VMs. While I suspect it may be a BIOS issue, I'd like to share in case the collective wisdom can guide me forward.
I created a VirtualBox jail using the template. It worked a dream and I tip my hat to the creators!
I can create virtual machines through the web gui, such as a "Other Linux 64-bit" machine for a CentOS 7 installation. I add the appropriate iso image on the CD/DVD virtual drive (CentOS-7-x86_64-Minimal.iso in my case). I start the VM and connect to it using TightVNC. The install starts, but almost always hangs within 5 seconds. If I remove the "quiet" from the kernel line in the install, I can see that the last message is "Switching to clocksource tsc". Screenshot attached.
I get the same issue with other distros (ubuntu-15.10-server-amd64.iso, debian-live-8.2.0-amd64-xfce-desktop.iso) and CentOS versions (CentOS-6.7-i386-minimal.iso). I have only tried Linux distros. Might not be limited to Linux, though.
I have checked that Hyperthreading is enabled in BIOS. I have also set BIOS to use Legacy, not UEFI BIOS as this has been noted as a potential issue for Linux installs.
I tried many, many different kernel commands (clocksource=hpet, noapci, intel_pstate=disable etc.). But the commands that were used on a successful boot, did not work on a new VM and they did not enable me to restart the VMs where CentOS had been successfully installed.
I have explored creating VMs with more RAM (up to 4096 GB), tried various System parameter settings (Nested paging enabled/disabled etc.) in the hope that I find a mismatch in settings that, once corrected, the VMs would successfully boot, but no joy.
The strange part is that I have managed to get two VMs to successfully boot and to complete the CentOS7 minimal installs, also installing GNOME GUI and Crashplan. (Details here if you're interested. Post #10). Rebooting the CentOS server within the VM worked fine, but as soon as I try to restart the VM it will not restart, hanging on the same "Switching to clocksource tsc" message. The VM settings and BIOS were unchanged. (I know because I gave up repeatedly restarting at 3 AM and just clicked "start" the next morning and was surprised to see it booting). If I had to guess, my VMs are successfully booting the image in 1% of cases.
I have noticed that the clock frequency on the line before the "Switching to clocksource tsc" ("Refined TSC clocksource calibration: 3356.699 GHz" in the attached example) is different every time. I assume this is because the frequency is being modified in line with performance needs of my FreeNAS server. Could it be that the VMs are only booting when a clock frequency is matching an expected value? That's how it feels.
I have contacted Asus for support to see what BIOS settings should allow Linux to be installed. So far all I get is "We do not support Linux OS...". I am pushing back, but I'm not holding my breath.
I asked in the VirtualBox forums. The bottom line there: "Since you are running *BSD you would get better help from them since this is their build and not ours". So no help there either.
I know VirtualBox is for advanced users and I'm not stupid enough to call myself that. ;) But getting this to work does appear to be the only way to get Crashplan to work. If I can't create any Linux installs in a VirtualBox, then I am left with solutions that run in FreeNAS jails or FreeNAS plugins, which is a far more limited set of options.
Is anyone else experiencing something similar when booting VMs in VirtualBox on FreeNAS? Any ideas what could be causing such behaviour?
I created a VirtualBox jail using the template. It worked a dream and I tip my hat to the creators!
I can create virtual machines through the web gui, such as a "Other Linux 64-bit" machine for a CentOS 7 installation. I add the appropriate iso image on the CD/DVD virtual drive (CentOS-7-x86_64-Minimal.iso in my case). I start the VM and connect to it using TightVNC. The install starts, but almost always hangs within 5 seconds. If I remove the "quiet" from the kernel line in the install, I can see that the last message is "Switching to clocksource tsc". Screenshot attached.
I get the same issue with other distros (ubuntu-15.10-server-amd64.iso, debian-live-8.2.0-amd64-xfce-desktop.iso) and CentOS versions (CentOS-6.7-i386-minimal.iso). I have only tried Linux distros. Might not be limited to Linux, though.
I have checked that Hyperthreading is enabled in BIOS. I have also set BIOS to use Legacy, not UEFI BIOS as this has been noted as a potential issue for Linux installs.
I tried many, many different kernel commands (clocksource=hpet, noapci, intel_pstate=disable etc.). But the commands that were used on a successful boot, did not work on a new VM and they did not enable me to restart the VMs where CentOS had been successfully installed.
I have explored creating VMs with more RAM (up to 4096 GB), tried various System parameter settings (Nested paging enabled/disabled etc.) in the hope that I find a mismatch in settings that, once corrected, the VMs would successfully boot, but no joy.
The strange part is that I have managed to get two VMs to successfully boot and to complete the CentOS7 minimal installs, also installing GNOME GUI and Crashplan. (Details here if you're interested. Post #10). Rebooting the CentOS server within the VM worked fine, but as soon as I try to restart the VM it will not restart, hanging on the same "Switching to clocksource tsc" message. The VM settings and BIOS were unchanged. (I know because I gave up repeatedly restarting at 3 AM and just clicked "start" the next morning and was surprised to see it booting). If I had to guess, my VMs are successfully booting the image in 1% of cases.
I have noticed that the clock frequency on the line before the "Switching to clocksource tsc" ("Refined TSC clocksource calibration: 3356.699 GHz" in the attached example) is different every time. I assume this is because the frequency is being modified in line with performance needs of my FreeNAS server. Could it be that the VMs are only booting when a clock frequency is matching an expected value? That's how it feels.
I have contacted Asus for support to see what BIOS settings should allow Linux to be installed. So far all I get is "We do not support Linux OS...". I am pushing back, but I'm not holding my breath.
I asked in the VirtualBox forums. The bottom line there: "Since you are running *BSD you would get better help from them since this is their build and not ours". So no help there either.
I know VirtualBox is for advanced users and I'm not stupid enough to call myself that. ;) But getting this to work does appear to be the only way to get Crashplan to work. If I can't create any Linux installs in a VirtualBox, then I am left with solutions that run in FreeNAS jails or FreeNAS plugins, which is a far more limited set of options.
Is anyone else experiencing something similar when booting VMs in VirtualBox on FreeNAS? Any ideas what could be causing such behaviour?