Critical Error?!

[MrMe01]

I've just been informed that my VM running Gentoo has suffered a critical error whilst emerging software. The problem is, it's still doing stuff. I can (or could, until the screen blanked) see the emerge progressing. Now I have to reboot it.

Why has this happened? How can I close this error message and leave the VM to continue? This should not have happened. I'm annoyed at this.

The log has been edited to make it small enough to upload, see attached.

[MrMe01]

Luckily, emerge is quite resilient to such things. Rerunning the last command before it failed seems to be working just fine.

[mpack]

07:41:05.412346 supR3HardenedErrorV: supR3HardenedScreenImage/NtCreateSection: rc=VERR_LDR_ARCH_MISMATCH fImage=1 fProtect=0xf0005 fAccess=0x2 \Device\HarddiskVolume2\Program Files (x86)\Google\Chrome\Application\chrome.exe:

Archicture mismatch: I think you saved the VM state, and then tried to restore it on a very different host.

If you want to move the VM around, don't use saved states. Use "Discard saved state" to get this VM running: it will act as if the PC had crashed.

[MrMe01]

07:41:05.412346 supR3HardenedErrorV: supR3HardenedScreenImage/NtCreateSection: rc=VERR_LDR_ARCH_MISMATCH fImage=1 fProtect=0xf0005 fAccess=0x2 \Device\HarddiskVolume2\Program Files (x86)\Google\Chrome\Application\chrome.exe:

Archicture mismatch: I think you saved the VM state, and then tried to restore it on a very different host.

If you want to move the VM around, don't use saved states. Use "Discard saved state" to get this VM running: it will act as if the PC had crashed.

I was doing absolutely nothing with the VM, besides click in the window to see the state of it as the screen blanks.

[loukingjr]

Well for one thing you are in the wrong section. You are on a Windows host, not Linux.

However, chrome.exe will never run in any Linux guest. It's a Windows program.

[MrMe01]

Well for one thing you are in the wrong section. You are on a Windows host, not Linux.

However, chrome.exe will never run in any Linux guest. It's a Windows program.

Well, I obviously clicked on the wrong subforum. Frightfully sorry, you know, it's not like an admin approved my post or anything...

If I had done something to manipulate the state of the VM, like snapshots or cloning or moving the VM in anyway that can't be done whilst the VM is running, why would it not first check to see if the VM is running and not attempt the instruction and cause the VM to crash? I'm not saying I tried anything like that, that would be stupid, especially as the VM is destined for a production environment.

I'll try not to post in the wrong subforum again if I ever visit this forum again.

[mpack]

Moved to "Linux Guests".

[MrMe01]

07:41:05.412346 supR3HardenedErrorV: supR3HardenedScreenImage/NtCreateSection: rc=VERR_LDR_ARCH_MISMATCH fImage=1 fProtect=0xf0005 fAccess=0x2 \Device\HarddiskVolume2\Program Files (x86)\Google\Chrome\Application\chrome.exe:

Archicture mismatch: I think you saved the VM state, and then tried to restore it on a very different host.

If you want to move the VM around, don't use saved states. Use "Discard saved state" to get this VM running: it will act as if the PC had crashed.

Here's the thing, whilst my interaction with the VM was blocked with a dialog telling me it had crashed, the console that I had open in the VM (tty1) was showing the usual activity that accompanies a typical emerge. Also, process hacker (Google it, very useful Windows program), showed activity on the VHD. The machine was still running, and running fine. It had full disk access. This is an issue with the governing process, not the VM process.

[Perryg]

Let's put this a little more in context. The way I see it the error shown was not in the guest but rather in the host.

Code: Select all

07:38:20.829804 VirtualBoxClient: detected unresponsive VBoxSVC (rc=RPC_S_SERVER_UNAVAILABLE 0x800706BA)
07:39:03.344557 VirtualBoxClient: detected working VBoxSVC (rc=ERROR_SUCCESS)
07:41:05.412346 supR3HardenedErrorV: supR3HardenedScreenImage/NtCreateSection: rc=VERR_LDR_ARCH_MISMATCH fImage=1 fProtect=0xf0005 fAccess=0x2 \Device\HarddiskVolume2\Program Files (x86)\Google\Chrome\Application\chrome.exe: 

But this might be a red herring. as there are multiple issues previous to this final error.

Code: Select all

07:33:55.563760 AHCI#0: Port 0 reset
07:33:56.082289 AHCI#0P0: Write request was active for 30 seconds
07:33:56.082313 AHCI#0P0: Canceled write at offset 3764609024 (319488 bytes left) returned rc=VINF_SUCCESS

This indicates a possible load on the host that is preventing the expedient transfer of data and causing a lag.

Possible cause is excessive CPU activity or drive speed lag due to IO caching.
I do see that the host has 2 cores 4 threads and that both cores are assigned to the guest.

Code: Select all

00:00:09.043562   NumCPUs           <integer> = 0x0000000000000002 (2)

00:00:12.709148 CPUM: Logical host processors: 4 present, 4 max, 4 online, online mask: 000000000000000f
00:00:12.709729 CPUM: Physical host cores: 2

VirtualBox can only deal with cores and not threads so I would suggest reducing the number of cores to 1 in the guest especially if the host is to be used for anything that requires heavy processing as this will cause issues for the guest.

You can also test the IO cache settings for the guest in the storage section as this might improve the performance, but I would also warn you to make a backup of the guest before trying anything if it is important to you.

Another thing I would watch out for is running a process explorer when not needed as it doubles the amount of load due to the way it works.

Finally while VirtualBox supports the VHD format it is by far the weakest of all the formats that you can use. VHD writes the headers but also has a footer that if improperly shut down can cause the guest to become corrupt making it useless. Of course this is your decisions but make sure you do timely backups.

[MrMe01]

Let's put this a little more in context. The way I see it the error shown was not in the guest but rather in the host.

Code: Select all

07:38:20.829804 VirtualBoxClient: detected unresponsive VBoxSVC (rc=RPC_S_SERVER_UNAVAILABLE 0x800706BA)
07:39:03.344557 VirtualBoxClient: detected working VBoxSVC (rc=ERROR_SUCCESS)
07:41:05.412346 supR3HardenedErrorV: supR3HardenedScreenImage/NtCreateSection: rc=VERR_LDR_ARCH_MISMATCH fImage=1 fProtect=0xf0005 fAccess=0x2 \Device\HarddiskVolume2\Program Files (x86)\Google\Chrome\Application\chrome.exe: 

But this might be a red herring. as there are multiple issues previous to this final error.

Code: Select all

07:33:55.563760 AHCI#0: Port 0 reset
07:33:56.082289 AHCI#0P0: Write request was active for 30 seconds
07:33:56.082313 AHCI#0P0: Canceled write at offset 3764609024 (319488 bytes left) returned rc=VINF_SUCCESS

This indicates a possible load on the host that is preventing the expedient transfer of data and causing a lag.

Possible cause is excessive CPU activity or drive speed lag due to IO caching.
I do see that the host has 2 cores 4 threads and that both cores are assigned to the guest.

Code: Select all

00:00:09.043562   NumCPUs           <integer> = 0x0000000000000002 (2)

00:00:12.709148 CPUM: Logical host processors: 4 present, 4 max, 4 online, online mask: 000000000000000f
00:00:12.709729 CPUM: Physical host cores: 2

VirtualBox can only deal with cores and not threads so I would suggest reducing the number of cores to 1 in the guest especially if the host is to be used for anything that requires heavy processing as this will cause issues for the guest.

You can also test the IO cache settings for the guest in the storage section as this might improve the performance, but I would also warn you to make a backup of the guest before trying anything if it is important to you.

Another thing I would watch out for is running a process explorer when not needed as it doubles the amount of load due to the way it works.

Finally while VirtualBox supports the VHD format it is by far the weakest of all the formats that you can use. VHD writes the headers but also has a footer that if improperly shut down can cause the guest to become corrupt making it useless. Of course this is your decisions but make sure you do timely backups.

The machine wasn't running brilliantly. I normally assign just a core to a VM, but as I'm dealing with Gentoo and all of the compiling that goes along with it, I thought that It would assign a core and a thread (used to VMWare, where all the cores, real or not are shown as cores)

I use VHD as I need to attach the drive directly to copy a large amount of data onto the VM (ext2fsd), all of the other formats are a pain to deal with when it comes to mounting the drive in Windows.

I only used the process explorer to determine if the VHD was still being accessed.


So, due to the way Virtualbox manages processing cores, I've assigned all of my processor time to the VM, not half, like I thought? If this is the case and the host becomes bogged down, why didn't it pause the guest until the load was reduced? Surely, due to the nature of the program, processor load would be monitored?

[Perryg]

If it only worked that way, but it doesn't. Monitoring across host/guest process would cost more than running the process explorer. At this point in time it is just the way it is sorry to say.

[Perryg]

Oh and just a little about threads. It is really a poor man's way of emulating cores. The threads share clock ticks with the same core in essence making it look like you have twice as much processor power when in fact it is not anywhere near that.

You mentioned compiling. This is the most processor intense operation that you can do with a virtual machine. I do compiling and I also do video rendering and the compiling is the heaviest. I also find notebooks just not up to the task. Yes it will work but you better walk away and have lunch while it is working. If you try to use the host or the guest while the process is going you can and often do cause issues.

[MrMe01]

If it only worked that way, but it doesn't. Monitoring across host/guest process would cost more than running the process explorer. At this point in time it is just the way it is sorry to say.

Luckily, Portage is a resilient nut and restarting the VM and rerunning the command worked without a single issue. It's still running now, under the same conditions just fine. I can only assume that both the guest and host became bogged down at the same time.

Windows was running slow, as was Chrome, but it wasn't cripplingly slow, it must've hit the VM more than any other process. Now I know more about the inner workings of Virtualbox, and its limitations, I'll be a bit more lenient as it is a brilliant piece of software for what it does.

I'll keep an eye on what I'm doing until I can poweroff the VM.

Thanks for your help here, and I do apologise for my rather snippy attitude earlier, loukingjr.

[Perryg]

Just another FYI. I noticed that you are running 5.0.0 and that is the first of the 5.* series and as with all software is prone to issues until they find what users can only provide to the DEVs to fix. I suggest that you upgrade to 5.0.8 as there were a lot of issues fixed.

Lastly look at the log. You are assigning almost all of the remaining memory to the guest.

• 00:00:08.024131 Host RAM: 3893MB total, 1283MB available
  00:00:09.043569 RamSize <integer> = 0x0000000040000000 (1 073 741 824, 1 024 MB)

This will not be an issue if you do not use the host for anything else but should you try to open something like chrome which uses a lot of CPU and an ample amount of RAM you will cause swapping and that might have been what started all of this to happen to you.

Code: Select all

07:38:20.829804 VirtualBoxClient: detected unresponsive VBoxSVC (rc=RPC_S_SERVER_UNAVAILABLE 0x800706BA)

[MrMe01]

Just another FYI. I noticed that you are running 5.0.0 and that is the first of the 5.* series and as with all software is prone to issues until they find what users can only provide to the DEVs to fix. I suggest that you upgrade to 5.0.8 as there were a lot of issues fixed.

Lastly look at the log. You are assigning almost all of the remaining memory to the guest.

• 00:00:08.024131 Host RAM: 3893MB total, 1283MB available
  00:00:09.043569 RamSize <integer> = 0x0000000040000000 (1 073 741 824, 1 024 MB)

This will not be an issue if you do not use the host for anything else but should you try to open something like chrome which uses a lot of CPU and an ample amount of RAM you will cause swapping and that might have been what started all of this to happen to you.

Code: Select all

07:38:20.829804 VirtualBoxClient: detected unresponsive VBoxSVC (rc=RPC_S_SERVER_UNAVAILABLE 0x800706BA)

I've assigned 1/4 of available RAM to the guest, I would have given more, but as I'm using the host for other tasks, I thought that would be enough. It's still compiling, using roughly 90 MB.

I'm awful for keeping software up to date (as it's usually a really clunky system like Java, where it just cannot do it in the background, it *has* to tell you it's updating), or on program launch, where I just want to use the program..