Courtney, Phil wrote:
I thought we might be on to something regarding the PGP disk encryption
software. Andrea Sormanni had also suggested disabling the encryption.
I moved my vdi to an external USB hard drive without encryption. I ran
for about 6 hours on Friday. However, my CentOS VM died today after
about 2.5 hours.
I got an Eclipse segmentation fault on my CentOS VM. An error dialog was
displayed stating "An error has occurred during virtual machine
execution." and "VDI: error reading pre-header in E:\CentOS.vdi." I had
to close the VM and upon reboot had numerous directory errors.
This is very mysterious. A failure to read the pre-header is a clear
host OS misbehavior, or an indication of a completely truncated image
(less likely). I see no reason why any OS would fail to read the first
72 bytes of a VDI file.
At the same time as the seg fault on the guest, I got "Insufficient
resources to complete request" errors on my Windows XP host.
Again this looks like the host OS has some serious issues.
At the time the error occurred, I wasn't doing anything on my guest but
I was doing a svn update of a very large repository on my host. Prior to
the failure, I had done socket communication between the guest and the
host over a bridged adapter.
I have uploaded the VirtualBox log file to
http://plc.pastebin.com/m1848894e.
The log file is pretty explicit (and repeats it to add more emphasis):
02:32:59.468 PIIX3 ATA: LUN#0: disk read error (rc=VERR_NO_MEMORY
iSector=0xd5f88d cSectors=0x8)
VirtualBox tries to get a little more memory, and the host OS rejects.
I'd say that you probably should be happy that VirtualBox handles this
rather gracefully.
The only solution seems to either avoid running other extremely
memory-intensive applications while VirtualBox is running as well, or
add more memory to the system. The VM isn't extremely big (512M, in a
system which has a total of 2G), but you clearly run out of memory.
I haven't tried using a PIIX4 controller type or a LSILogic SCSI
controller yet (as suggested by Andrea).
That makes no difference if the host misbehaves reading the VDI file or
is completely out of memory...
Klaus
Phil
------------------------------------------------------------------------
*From:* Klaus Espenlaub [mailto:[email protected]]
*Sent:* Fri 1/29/2010 12:01 PM
*To:* Courtney, Phil
*Cc:* [email protected]
*Subject:* Re: [vbox-users] Linux Guest IDE Read and Write Errors
Hi Phil,
Courtney, Phil wrote:
> This is my first time for submitting an email. So please let me know if
> I should do something differently, include additional information, etc.
I wish we'd get more such problem descriptions - all relevant details
included. It's be totally perfect if you'd have included a reference to
the VBox.log file (menu: Machine->Show Logs..., uploaded to some
pastebin server), because that'd have saved you some breath and provided
even more detail.
> I am getting IDE disk read and write errors. I have not been able to
> associate these errors with any particular activity on the guest or the
> host except that it seems like the errors are more likely to occur after
> I have been communicating over sockets between the guest and the host. I
> was running with both a bridged adapter and a NAT adapter. It seems to
> be more stable (i.e. will run longer) after removing the NAT adapter. I
> don’t know if the networking has anything to do with the problem or is
> just a coincidence. However, after seeing the errors on the guest, the
> host will also exhibit errors such as “Insufficient resources to
> complete the request” when trying to copy files or an inability to start
> new applications. I usually need to restart my host machine in order to
> clear these problems.
That's a bit unclear - maybe the log would show hints what's going wrong
if networking is added to the picture.
> The errors occur somewhat randomly. I am usually able to run for 1 to as
> many as 6 hours before experiencing a problem. Occasionally I can run
> for a few hours and shut down without a problem. Once I get an error, it
> will repeat continuously until I power off the VM. I cannot do a normal
> shutdown.
It's unusual that the problem you reported persists (see below).
> I am seeing “PIIX3 ATA: LUN#0: disk read error” and “PIIX3 ATA: LUN#0:
> disk write error” messages in the VirtualBox logs. I also see “Failed
> opcode was: unknown”, “hda: read_int: status = 0x41 {Drive Ready Error}
> error = 0x10 {SectorIdNotFound}”, and “EXT3-fs error in
> start_transaction: Journal has aborted” errors displayed in the VM
window.
>
>
>
> Upon restarting the VM, I do a file system integrity check. It will
> sometimes pass without a problem. However it often finds various errors
> including invalid attribute and inode reference counts, unattached
> inodes, block bitmap differences, free blocks count wrong, and free
> inodes count wrong. So far I have selected to fix these errors and the
> VM will then boot successfully.
>
>
>
> I have run chkdsk on the host several times (before and after fixing the
> vdi errors on rebooting the VM) and have not had any errors reported.
>
>
>
> My host configuration is:
>
>
>
> Windows XP Service Pack 2
>
> Intel Core2 Duo CPU – T9600 at 2.80 GHz
>
[...]
>
>
>
> Additionally, I am running PGPDesktop disk encryption on the SATA hard
> drive. I am also running the Cisco VPN Client in my host machine, but I
> am not connected to the VPN when communicating between the guest and
host.
We had reports for other encryption software that it results in widely
fluctuating write speeds, from near zero delay to many seconds.
Those periods of really slow response time to writes are critical - some
really picky guests allow a maximum of 10 seconds for a write command to
complete once the the transfer to the device has begun. Others are less
impatient, but generally it's not possible to meet the guest OS
requirements if the disk image lives on an encrypted disk.
You have a dual core system, so there's not much room if both encryption
and the VM each fully utilize a core.
> VirtualBox and Virtual Machine configuration:
>
>
>
> VirtualBox Version 3.1.2 r56127
>
> CentOs Version 4.3 OS, Kernel version xxxx
>
[...]
>
> If anyone has a clue as to how to diagnose or fix this problem, it would
> be greatly appreciated.
Verifying my hypothesis is easy: move the disk image to a non-encrypted
drive. Even an external USB disk should do.
Klaus
------------------------------------------------------------------------
_______________________________________________
vbox-users mailing list
[email protected]
http://vbox.innotek.de/mailman/listinfo/vbox-users
--
Dr. Klaus Espenlaub
Sun Microsystems GmbH
Werkstrasse 24
71384 Weinstadt
Germany
=====================================================================
Sitz der Gesellschaft:
Sun Microsystems GmbH, Sonnenallee 1, D-85551 Kirchheim-Heimstetten
Amtsgericht Muenchen: HRB 161028
Geschaeftsfuehrer: Thomas Schroeder, Wolfgang Engels, Wolf Frenkel
Vorsitzender des Aufsichtsrates: Martin Haering
=====================================================================
_______________________________________________
vbox-users mailing list
[email protected]
http://vbox.innotek.de/mailman/listinfo/vbox-users
