VMware Communities
ddoherty
Contributor
Contributor

Fusion Beta 1.1 Hangs; NTFS corruption?

Hi,

I'm running 1.1 Beta on a MacBook, using a Boot Camp partition. WinXP. Running Tiger on the Mac. 2GB memory -- 1 each to Mac OS X and WinXP.

Running Fusion full-screen in a desktop managed by Desktop Manager (third party app).

Three times over the last couple of weeks, the following has happened:

1. The VM partially locks up -- I can't get to any other Mac applications. The cursor turns into a text "I-bar." I can choose to shut down the OS in the Start menu, but the OS shuts down only part way, then hangs. Have to power cycle. Mac OS "force quit" key combinations don't work. Ctrl-Alt-Delete does nothing in Windows.

2. At the time, I have exported (shared) some directories on my XP (Boot camp) file system to Mac OS. And I have had shared files opened in BBEdit.

3. When get things back up, Windows tells me that the NTFS is corrupt and needs to run chkdsk. I run this and always lose files -- typically the ones that

were open in BBEdit. Sometimes the directory containing the files is corrupted.

Note that this always happens when I am working inside the VM, not doing something on the Mac OS side.

I saw a cryptic note in the bugs section of the release notes for an earlier Beta about possible Boot Camp file corruption in "rare cases" (or somesuch wording...). Is this a cas of this?

Anyone else seen this? Or have any advice?

Thanks.

Tags (3)
Reply
0 Kudos
16 Replies
kvishwa
Contributor
Contributor

I have the exact same issue but now when I open my VM image (XP Pro), I get this error - File "/Users/Karthik/Documents/Virtual Machines.localized/Local Windows XP Pro.vmwarevm/Local Windows XP Pro.vmx" line 62: Syntax error.

Does anyone know how to get this error resolved. Looks like my vmx file is missing.

Any advise to enable me not reinstall the entire image will be helpful.

Thanks... K

Reply
0 Kudos
chadseld
Contributor
Contributor

I'm having the same hanging problem. I'm running 1.1 final. VMWare hangs. Force quitting VMWare leaves the vmware-vmx process running. The VMWare icon remains in the dock. I have to restart my mac in order to run VMWare again. This is very annoying. I am running 2 monitors, I wonder if this is related to the problem?

Reply
0 Kudos
chadseld
Contributor
Contributor

For the archive... I have a MacBook Pro with 3GB RAM, 512 allocated to VMWare. I am running 2 monitors (laptop + external). I am running with experimental DirectX support turned ON and the number of virtual CPU's set to 2. The virtual OS is Windows XP SP2.

I'll try running with DirectX off and CPU's to 1 and see if the problems reoccur.

The hanging usually occurs while I'm using a mac application and VMWare is in the background. Thus, I only notice the hang when trying to switch back to my windows environment.

Once vmware is hung, if I kill VMWare.app and vmware-vmx, I can log-out and log-in again and re-launch VMWare.app. When trying to launch my windows environment I get an error stating "This virtual machine appears to be in use." If I click the "Take Ownership" button, there is no effect. A complete restart of Mac OS X is required to get VMWare working again.

Restarting Mac OS X is a problem because once VMWare has locked up in this way, Mac OS X hangs on restart. It logs out, but fails to completely shut down and restart. I have to hold down the power button and force the machine to restart.

Reply
0 Kudos
HPReg
VMware Employee
VMware Employee

This sounds like a separate issue. Please send the content of the .vmx file. Have you edited this file by hand with TextEdit by any chance?

Reply
0 Kudos
HPReg
VMware Employee
VMware Employee

When "VMware hangs", do you see any line containing the word "vmmon" in /var/log/system.log appear?

Reply
0 Kudos
kvishwa
Contributor
Contributor

No. I have not edited anything. Am not sure what to look for. If these are not the ones. Am not able to find a .vmx for this image.

Please let me know if there is something I am missing here, please let me know. I have 4GB RAM on my machine and 1 GB allocated to this image. And 1.5 GB allocated to another. The second image - Windows Server 2003 works just fine.

Thanks for this help.

Appreciate it.

Reply
0 Kudos
ddoherty
Contributor
Contributor

Sorry to try to bring this back on topic... Smiley Wink

But I see file system corruption as a pretty major problem.

Sure, I'd like to see the hangs fixed, but if they results in open files and directories being corrupted, that's kind of a show-stopper for me.

Is vmware looking at this? Is this the same problem as the cryptic "file system corruption in Boot Camp partitions" that I saw under the release notes at some point?

Reply
0 Kudos
HPReg
VMware Employee
VMware Employee

ddoherty,

When you say you are sharing your NTFS partition with Mac OS, how do you do it exactly?

The fact that your NTFS partition is corrupted, while annoying, is explained by the fact that the VM suddenly dies: it is no different from pulling the plug on a physical computer.

So the real problem to solve here is to understand why Fusion hangs. And for that I need more info from you. Read my previous post.

Reply
0 Kudos
ddoherty
Contributor
Contributor

I'm not a file system guru, but it doesn't seem right to me that, if the OS (guest OS - WinXP) hangs and the machine is power-cycled, files that were open (though there was no i/o currently being done to them) and directories (folders, inodes ?? -- I'm not familiar with the NTFS lingo) would be corrupted.

Here's the situation (hopefully expands on what was said in the inital post):

Running Fusion 1.1 FC (at the time, now 1.1 final).

Some folders from NTFS (Boot Camp partition) have been shared from WinXP. These have been mounted in Mac OS X (Tiger) using the Go menu >> Connect to Server.

I had shared files open on the Mac side, in BBEdit.

Running Fusion in Full Screen mode. Working in WinXP (in Fusion). Fusion hangs...

The cursor turns into an I-bar and I can't get to any other Mac Aps or the Finder. Can choose shutdown in the Start menu, which partly shuts down WinXP, then hangs.

Power cycle the machine, and chkdsk reports corrupted files and directories, These often include the files that were open in BBEdit. But sometimes, they are files there were RECENTLY edited but were not open at the time. The corrupted directories are those that contain these files. It's possible that the corruption occured before Fusion/XPbecame hung. It's also possible that the file system problem is related to the hang, for that matter.

I usually can't force any of the Mac OS apps to quit (e.g., via ctrl-cmd-esc), but sometimes can.

(perhaps) Important note: At the time I was running Desktop Manager (http://desktopmanager.berlios.de/), with Fusion running on one desktop and BBEdit running on another. I have since been running without Desktop Manager for about a week and haven't seen hangs or file corruption.

From the release notes for 1.1 final:

  • Using the Boot Camp partition as a virtual machine can, in a few instances, result in a corrupted Boot Camp partition.
    There currently is no known workaround. Work is in progress on this problem.

Is this the same problem that I am seeing?

I may be wrong, but it seems wrong that a system hang can, for instance, corrupt a directory that contains a file that is open (via file sharing), and to which there is no active i/o going on. This may be a pure XP/NTFS issue and not a Fusion issue, but I'm not convinced.

I do agree that fixing the causes of the hangs is important, but if there is something else wrong with the file system implementation via Fusion, that's even more important.

And no, I'm not going to start running Desktop Manager again to see if this is the cause of the hangs. I'm in the middle of preparing a big software release and can't afford the (cycle of chkdsk, lose files, reload from backups) loss of productivity.

Thanks for looking into this.

Reply
0 Kudos
HPReg
VMware Employee
VMware Employee

I'm not a file system guru, but it doesn't seem right to me that, if the OS (guest OS - WinXP) hangs and the machine is power-cycled, files that were open (though there was no i/o currently being done to them) and directories (folders, inodes ?? -- I'm ot familiar with the NTFS lingo) would be corrupted. I may be wrong, but it seems wrong that a system hang can, for instance, corrupt a directory that contains a file that is open (via file sharing), and to which there is no active i/o going on. This may be a pure XP/NTFS issue and not a Fusion issue, but I'm not convinced.

On modern OSes, when you modify a file, the modification goes to the cache and it is not immediately flushed to disk. So if a machine hang occurs in between, your modification (the data), and sometimes other things (filesystem metadata like time of last access) are lost. NTFS is a journalling filesystem, which means it tries to mitigate that problem by using a transactional log. When you say corruption occurs, it is corruption that can be repaired with NTFS's transaction log or not?

Some folders from NTFS (Boot Camp partition) have been shared from WinXP. These have been mounted in Mac OS X (Tiger) using the Go menu Connect to Server.

OK, this is what I wanted to clarify. Your are using network shares. Good. You are not mounting your NTFS partition read/write on Mac OS. Unexpected simultaneous access to the partition from both Windows and Mac OS would certainly lead to NTFS filesystem corruption.

Working in WinXP (in Fusion). Fusion hangs...

You still haven't told me if, when Fusion hangs, you see a Panic line from vmmon in /var/log/system.log Smiley Happy

What hangs exactly? Apparently you can still interact with the Fusion UI. IS it just the display of the VM which appears hung?

I have since been running without Desktop Manager for about a week and haven't seen hangs or file corruption.

That is why I believe the corruption is related to the hang.

Using the Boot Camp partition as a virtual machine can, in a few instances, result in a corrupted Boot Camp partition. Is this the same problem that I am seeing?

I don't think so. The corruption we were referring to in that release not is one that prevents you from even booting Windows.

if there is something else wrong with the file system implementation via Fusion, that's even more important.

There is no such thing as a file system implementation in Fusion. Fusion gets a request from the guest OS (which is the one who implements file system code) to read a sector from the virtual hard drive, and Fusion turns around and asks the host OS (here Mac OS) to read the same sector on the physical partition. If the sector mapping done by Fusion was incorrect, your guest OS would not even boot.

And no, I'm not going to start running Desktop Manager again to see if this is the cause of the hangs. I'm in the middle of preparing a big software release and can't afford the (cycle of chkdsk, lose files,reload from backups) loss of productivity.

I understand and I'm sorry our software hangs like it does in your case.

ddoherty
Contributor
Contributor

>You still haven't told me if, when Fusion hangs, you see a Panic line from vmmon in /var/log/system.log

No, I didn't check this. If it happens again, you can be sure that it'll be the first thing that I look at. :smileyblush:

>When you say corruption occurs, it is corruption that can be repaired with NTFS's transaction log or not?

Dunno how to do this. All I know is that chkdsk ran and blew away files and directories. Nothing was recoverd. How does one use NTFS's transaction log?

>

What hangs exactly? Apparently you can still interact with the Fusion UI. IS it just the display of the VM which appears hung?

I can't interact with the Fusion UI. Remember that I'm in full screen mode. When it hangs, I can't, e.g., move the cursor to the top of the screen to enable the menus. I can't ctrl-cmd-S to get out of full screen mode. I can't (usually) issue the ctrl-cmd-esc Mac three-fingered salute to force quit anything -- although sometimes this works.

I can still minimally minimally interact with Windows. The cursor turns into a standard text "I-bar" and I can use it to click the start menu and try to shut down. Shutdown appears to progress for a while, then the whole system (WinXP and Mac OS X) appears to hang. Then I power cycle.

>There is no such thing as a file system implementation in Fusion.

Fusion gets a request from the guest OS (which is the one who

implements file system code) to read a sector >from the virtual hard

drive, and Fusion turns around and asks the host OS (here Mac OS) to

read the same sector on the physical partition. If the sector mapping

done by Fusion >was incorrect, your guest OS would not even boot.

I didn't mean to suggest that Fusion implements the fs. I was suspicious of the "Fusion gets a request from the guest OS (which is the one who

implements file system code) to read a sector from the virtual hard

drive, and Fusion turns around and asks the host OS (here Mac OS) to

read the same sector on the physical partition" part.

I've worked in Windows NT and XP (and many flavors of Unix; and Mac OS since 1984) for a long time. Over those many years, many times I was hit by many hangs and power interruptions. I don't recall the predictable pattern of NTFS file corruption that I see here. That said, I'm not trying to argue, I'm just letting you know the gut feeling of someone who has been in this business a long time. I also realize the debugging such a not strictly reproducible problem is a... challenge. :smileydevil:

Regardless -- I am suspicious of the "Desktop Manager" as a potential source of the hangs.

Thanks for your responses. I do appreciate the straight talk. Let me know if I can provide any more info.

btw, I am very impressed at the thorough list of outstanding problems in the release notes. This is very useful and such transparency is to be applauded.

Dave D.

Reply
0 Kudos
Randomness
Contributor
Contributor

I just want to add to this thread that I'm seeing a similar hanging issue, but with Ubuntu 7.10 instead of Windows XP. I'm using Fusion 1.1 (final) on OS 10.5.1

Reply
0 Kudos
HPReg
VMware Employee
VMware Employee

No, I didn't check this. If it happens again, you can be sure that it'll be the first thing that I look at.

You can look at the file after the fact.

All I know is that chkdsk ran and blew away files and directories. Nothing was recoverd. How does one use NTFS's transaction log?

You do not have to do anything. That is just the way NTFS works. The NTFS transaction log should guarantee the integrity of at least your metadata (file and directory structure), so this surprises me just as it surprises you.

I can't interact with the Fusion UI.

Can you ssh to your hung Mac from another Mac, and do a ps auwwx and send me the output?

The cursor turns into a standard text "I-bar"

Windows style or Mac OS style?

ddoherty
Contributor
Contributor

>You can look at the file after the fact.

I looked after you mentioned it, but I didn't see anything that old. This hasn't happened in about 2 weeks.

>Can you ssh to your hung Mac from another Mac, and do a ps auwwx and send me the output?

I will try, but as I said in an earlier post, this hasn't happened to me since I stopped using Desktop Manager. And I'm in the middle of preparing a big software release and am not willing to (re-)tempt fate. As time goes on without more hangs and file corruption, I'm becoming more convinced of some kind of bad interaction with Desktop Manager. This would be too bad, because fusion and virtual desktops are a natural fit. I haven't yet tried Leopard and its virtual desktop (spaces?). Anyone heard if Fusion plays well with that?

>Windows style or Mac OS style? ("I-bar")

I didn't look too closely at the time, but I'm pretty sure that it was Windows-style (no cross-hatch in the middle, if that's what you mean).

If it does happen again, I will try to gather more data as you suggest.

thanks for your follow-up.

btw, how does one embed prior posts into a new post in this forum?

Reply
0 Kudos
djdevvydev
Contributor
Contributor

FWIW, I'm experiencing what seems to be the very same issue.

Whenever the host machine shuts down unexpectedly, for whatever reason -- a crash or a dead battery -- all open data files in XP under Fusion are corrupted irrecoverably. I'm running OS X 10.4.11 and Fusion 1.0 (51348). I agree with ddoherty: I've been working with XP systems for a long time and this corruption -- links broken on every open data file, even though no I/O had occurred for many minutes before the crash -- seems very unusual.

Incidentally I am running on a Boot Camp partition; I don't know if that matters?

I've lost many hours of work to this issue at this point. Starting to get pretty frustrating. I'd welcome any suggestions.

Thanks -

DJD

Reply
0 Kudos
jrwti
Contributor
Contributor

This is the exact same problem that has happened to me twice now. I am running Mac OSX Leopard, VMWare Fusion 1.1.1, Bootcamp 2.0.x (haven't upgraded to 2.1 yet), XP Pro x64 on a Mac Pro quad-core. In both occurences, a power outage has caused the computer to restart. Whatever files are open in XP at the time of the power outage are corrupted.

In my case, they still exist on my computer (in a corrupted form) until I run chkdsk, which inevitably deletes them. I have tried to run chkdsk /r , but it asks me if I want to run chkdsk on restart, and the restart promptly forgets the repair argument and deletes the corrupted files.

As with the previous posters, many of these files, though "open" by an application, have been saved, often for a long time with no I/O going on.

What the other posters haven't mentioned is whether or not they had their bootcamp partition mounted as a volume on their Mac's desktop. I know that my bootcamp volume was mounted, and read (after this problem occurred) that it is extremely dangerous to mount a guest filesystem on the host machine. I promptly umounted my bootcamp partition by ejecting it. It is not clear to me whether or not this affects the eventual outcome of files that get corrupted by simply being "open" without any disk I/O when the failure (power / crash /etc) occurs.

Like the others, in my extensive experience with native Windows and filesystems in general, I have never experienced this kind of corruption from a power outage. Certainly, in the course of standard Windows OS usage, if a power failure were to occur, I might lose whatever data I hadn't saved recently, or maybe a file would get corrupted if the power went out right as I was writing to it. Those are acceptable loss scenarios, the former being operator error, and the latter being understandable (and mostly recoverable). The kind of loss scenario I and others have described is not acceptable: major corruption, little to no chance of/opportunity for recovery, and the loss of the entire file (not just the last 20 minute's work). Obviously I am going to have to step up my saving/backing up procedure to the level of "what you do when you don't trust the operating system to maintain your files' integrity."

I was under the impression that since I was accessing an actual disk

partition via Boot Camp, that my files would be less susceptible to

corruption than if I were to create a Windows .vmdk and work from there. Had I known that this gremlin would rear its ugly head and cause me to lose days of work, perhaps I would have reconsidered using VMware Fusion at all. More specifically, Had I been aware of this issue before it happened to me, I would have changed my file management procedures to ensure minimal loss. Unfortunately for me and others, all we are left with is corrupted files and the question "why?"

Reply
0 Kudos