VirtualBox

Ticket #10702 (new defect)

Opened 22 months ago

Last modified 22 months ago

AHCI timeouts on heavy I/O

Reported by: Kohtala Owned by:
Priority: major Component: other
Version: VirtualBox 4.1.18 Keywords:
Cc: Guest type: Linux
Host type: Windows

Description

I tried compiling Ångström distribution on my i7 with four virtual CPU. AHCI starts to timeout during build and causes filesystem corruption for the guest. Host is Windows 7, guest is Ubuntu 10.04 LTS. I first saw this with VirtualBox 4.1.16, but upgraded to 4.1.18 and saw it on it as well.

bitbake conf/local.conf has these lines

PARALLEL_MAKE = "-j4" BB_NUMBER_THREADS = "2"

It gives the possibility of running two make -j4 in parallel.

I found bug #10292 that seems to relate to heavy disk I/O, but I've not seen an assert so I create this new ticket instead.

Attachments

dvsdk_dm3730-2012-06-24-16-50-42.log Download (129.8 KB) - added by Kohtala 22 months ago.
vbox.log
sata-err-2012-06-24-dmesg.txt Download (92.7 KB) - added by Kohtala 22 months ago.
dmesg from guest

Change History

Changed 22 months ago by Kohtala

vbox.log

Changed 22 months ago by Kohtala

dmesg from guest

comment:1 Changed 22 months ago by frank

There is probably not much we can do about the AHCI timeouts, it just means that your host hard disk is not fast enough serving all disk requests. You could also try if it makes any difference if you enable the host cache for the AHCI controller (VM settings).

The guest disk corruption is more relevant. Are you sure that the disk corruption happens with VirtualBox 4.1.18 as well or could it happen that you used your VBox 4.1.16 disk image which was already corrupted to a certain point with VBox 4.1.18? I'm asking because 4.1.18 contains a fix which could be relevant in your case. So I would prefer if you could start from a really sane guest image to try to reproduce this corruption.

comment:2 Changed 22 months ago by Kohtala

Ok. I looked around and found that it is the guest operating system that times out I/O.

I used on 4.1.18 the VDI from 4.1.16. I've since done VBoxManage clonehd and fsck. But it is rather time consuming to run into this problem and fix the problems caused. I have some work to do, so can not just play with this.

I moved to competing sw for running virtual machine and it used LSI MPT SCSI device without problems so far. It seems some other controller but SATA might be nicer on the host. I did get further using IDE, but then I installed a new Ubuntu 12.04 LTS and ran into some problems with that as well, but had no time to investigate (I may have corrupted it with SATA first before IDE problems, so could have been corrupt VDI or something).

I also consider adding to guest Linux file /etc/udev/rules.d/95-vbox-ioscheduler.rules with a line

DEVTYPE=="disk", ENV{ID_VENDOR}=="VBOX", ATTR{queue/scheduler}="deadline"

in an attempt to modify the I/O behavior of the guest.

Version 0, edited 22 months ago by Kohtala (next)
Note: See TracTickets for help on using tickets.

www.oracle.com
ContactPrivacy policyTerms of Use