Serious crash and disk corruption

Discussion in 'Parallels Desktop for Mac' started by elarsson, Apr 10, 2006.

  1. elarsson

    elarsson

    Messages:
    3
    I installed Parallels and a stock Windows XP SP2 virtual machine earlier today and everything was going quite well on my new MacBook Pro. When joining my Windows domain at the office, however, the Mac experiened a kernel panic, and then would no longer boot. Disk Utility was not able to recognize the boot volume as a valid HFS+ volume (it saw a partition, but it was unlabeled and it ouldn't find a catalog). It looks like the crash took out a signifiant portion of catalog and other essential strutures on the disk.

    Since this is a new MacBook Pro with no data on it, I opted to simply do a reinstall rather than try any further to recover data on the disk.

    Such are the perils of playing with beta software, eh? However, after years of playing with bleeding edge VMWare betas on my Windows and Linux machines and not experiening anything destrutive to the host system, I am a bit disturbed that this happned with Parallels. Perhaps things are not quite as isolated from the hardware as they shoulde be? I for one, would be willing to put up with a little performance degredation for some more safety...
     
  2. >>> Message has been deleted by the user <<<
     
  3. elarsson

    elarsson

    Messages:
    3
    Well, one can't get the panic.log off the system if the boot volume is destroyed.

    Firstly, I am not running this on a production machine with actual data on it (and even on a production machine, one certainly should keep good backups). That being said, rebuilding a test machine is more than a minor annoyance.

    Generally, when a product has reached the "beta" phase we expect that it has gone through some internal QA. Releasing beta software that has not gone through sufficient internal QA, where show stopping, file system corrupting bugs have not been reasonably weeded out is irresponsible. I am not expecting "perfection" as you so put, but I think it is reasonable to expect the software to be engineered for safety. On modern hardware, we can spare some performance for safer software.

    I am not the only one here who has expressed concerns about the design of this software. It is true that all virtualization products require code running in kernel space, but this should be kept to a minimum. Certainly, access to the disks should be delegated to the host operating system. Enterprise-class virtualization products like VMWare ESX Server mitigate the risks associated with a full-blown hypervisor layer by maintaining strict control over the underlying host or "console" OS, which is very stripped down, does not run applications, and is only there for the purposes of supporting the virtualization environment. On a desktop, one does not have this luxury of "owning" the host OS. Therefore, safer practices are indicated. Competing desktop virtualization products achieve the correct (imho) balance between providing direct access to the hardware (for speed) and delegating tasks to the host (for safety).

    I am also not the only one on the forum who has reported filesystem corruption. A simple crash of the OS does not do this. The journaling filessytem can recover from the journal if the system was uncleanly shutdown in mid transaction. That's not what we're seeing here. This is corrupt data being written to disk, and not just to the virtual-disk container file sandbox that Parallels is supposed to be writing to, but to other areas of the disk. As far as bugs go, that is pretty serious.. Something that should have been identified in internal QA.

    I'm not writing the product off, by any means, and will keep playing with it. It isn't the "unclean masses" that one need be concerned about though, it is the professionals who are forming impressions based upon this release. The expectations for beta are different than alpha. Parallels does themselves a diservices by calling the software in its present state "beta".
     
  4. PubGuy

    PubGuy

    Messages:
    119
    I've retried again, this time using a SPARSE image for PW

    I have a kernel crash that had required a 6 pass repair of my OS X volume. I agree that this should not be happening once you declare a BETA stage for testing.

    However, I really needed to have a few Windows-only programs operational on my Macbook and decided to try again. This time, I created a 4 GB sparse disk image in disk utility, mounted it, then have Parallels install the file and drive image into the sparse disk image. This way, if the PW causes a kernel panic while writing to disk, it will be writing to the disk inside the sparse disk image and not on my main volume. The only volume that should be corrupted at that point should be the disk image and not my main system.

    So far so good, have not had another kernel panic since, but I'm very careful about not going on-line while PW is running. I've made a copy of my fully configured disk image, just in case, so that I won't have to do a full Windows reinstall and configure again. The only thing that I do different now is to mount the disk image on my desktop first, then launch PW, and when I get done and quit PW, I then unmount the drive image.

    As I said, so far so good.
     
  5. >>> Message has been deleted by the user <<<
     
  6. elarsson

    elarsson

    Messages:
    3
    Putting the virtual disk inside a mounted .dmg is an interesting suggestion. I'll have to give that a shot and see how it goes...
     
  7. PubGuy

    PubGuy

    Messages:
    119
    Yeah, this came to me when my kernel panic occured and I started having weird behavior with some file. Disk utility indicated a bad volume. So it booted off the CD and let disk utility do its thing to fix it. Interestingly, I had 2 partitions on that disk, one of them was created by boot camp. The other volume was fine. Also, I am using a file vault for image for my home directory, and that was untouched also.

    So, theoretically, since the kernel panic occured and trashed the volume on which the virtual drive was residing on (and did not trash my file vault image), I figured that using a sparse disk image and installing the virtual drive into the disk image will contain any possible volume corruption to the mounted disk image. Once Parallels resolves these kernel panics, you can then copy your virtual disk image out of the sparse image file. Plus -- just in case -- once your virtual disk is fully configured, simply make a copy of the sparse disk image so you won't have to reinstall yet again.

    Anyways, based on my observations, this should help protect your primary system from corruption. :D
     
    Last edited: Apr 11, 2006
  8. loughkb@yahoo.com

    loughkb@yahoo.com

    Messages:
    44
    It's just crazy that it works..... I decided to play around a bit and go one further.

    I have a fileserver on my lan, running Linux with ~300gig of drive space. I read your thread and had a crazy idea...

    On my dual core mini, I have an SMB share mounted from the fileserver. On that share, I created a 5Gig sparse DMG image. On the mounted image, I had Parallels create it's 4Gig virtual drive.

    So we have, (((Parallels VM Virtual drive)DMG file)Samba Share)Fileserver drive.

    Does it work, yes!, is it slow as a turttle? No! In fact it's surprisingly quick since all access to the virtual drive is going across the ethernet.

    Ok, now, the why of this experiment. First, A VM like parallels give one the freedom to play around with ideas as easily as legos. Secondly, If it works well enough, and your a tinkerer, you could create many images on a fileserver and have easy access to many OS's. Third, how much more isolated can the virtual drive be? (grin) Oh, and you don't have to use any space on your system drive.

    To really test this bizarre configuration, I downloaded the latest Ubuntu installer ISO, saved it to another share on the fileserver and pointed the VM at it for the CDrom choice.
    Booted the VM, installed Ubuntu, rebooted and tested. Everything works quite fast, I've had to double check to make sure I was actually running off a network share.

    Kevin L. Computer Geek, Mad Scientist...
     
  9. Ben @ Parallels

    Ben @ Parallels

    Messages:
    56
    What we mean by Beta

    I think that this sums it up nicely. Parallels Workstation 2.1 for Mac OS X is a "true beta", meaning that there will inherently be bugs associated with it. Our goal with this beta release was to get a working version out to the public, and then work directly with the Mac community to make it stronger, faster, more powerful, and bug-free.

    As of now, we have more than 45,000+ beta testers helping with the process, and we receive thousands of constructive feedbacks every day. With your continued help - and patience - we'll have a rock solid, bug-free version of the product ready in no time!

    Remember, if you encounter an issue or have a suggestion, please submit it via our online feedback form, available at http://www.parallels.com/en/support/mac/.
     
  10. dhjdhj

    dhjdhj

    Messages:
    220
    How does one create such DMG images and mount them?
    Thanks,
    David Jameson[

    QUOTE=loughkb@yahoo.com].....I created a 5Gig sparse DMG image. On the mounted image, I had Parallels create it's 4Gig virtual drive.[/QUOTE]
     
  11. John Mac

    John Mac

    Messages:
    14
    Ever since installing Parallels (Beta 1 and 2) I have been experiencing disk corruption as identified (and fixed, luckily) by Disk Utility.

    I am now extremely nervous about launching the Parallels VM (running XP) for any extended period of time since it has been causing repeated filesytem corruption. I have not encountered any kernel panics either so this makes me even more nervous. I like the idea (and speed) of Parallels, but I can't afford to lose data on my MAC OS disk.

    Just my 2 cents. I really hope this gets fixed. Great software other than that.

    -- JM.
     
  12. >>> Message has been deleted by the user <<<
     

Share This Page