Opened 13 years ago
Closed 12 years ago
#9749 closed defect (fixed)
useHostIOCache does not automatically turn on for some vm's
| Reported by: | Spyplane | Owned by: | |
|---|---|---|---|
| Component: | other | Version: | VirtualBox 4.0.12 |
| Keywords: | Cc: | ||
| Guest type: | other | Host type: | Linux |
Description
I've had several guest vm's have severe filesystem corruptions. I found today that my problems might be due to useHostIOCache being set to False on all of my vm's. I only use vboxmanage, instead of the UI, so I don't get the warnings regarding ext4, linux and not having useHostIOCache enabled. After talking to 'aeichner' on IRC, he mentioned that vbox should be automatically enabling it in my version of virtualbox (4.0.12), but after checking my logs, this is not the case. I've attached the logs of one vm that has had inode errors and still has useHostIOCache set to false in the .vbox files.
I've had this issue on 6 identical servers running Ubuntu Server 11.04 64bit, with a variety of virtualbox vm's. I've had FS corruption on 15+ vm's.
Attachments (1)
Change History (6)
by , 13 years ago
follow-ups: 2 3 comment:1 by , 13 years ago
comment:2 by , 13 years ago
Replying to frank:
Your host kernel is 2.6.38 which does not have this problem (it was fixed in 2.6.36). So the question is, where your corruptions come from. Sorry for asking, I have to ask this: Did you always terminate the VMs properly or did some VMs crash which could have resulted in such a corruption? And does your host disk which is mounted at /vms1 behave correctly or does it show any errors?
Thanks for the info on the kernel version. I haven't upgrade my kernel, so these issues were seen on the kernel listed in the logs. All of the vm's where I had issues happened on running vm's with no prior problems. The dmesg logs would show inode errors (for the linux guests) and zfs errors (on my solaris / opensolaris guests) and in both cases the guest fs's would remount read-only. That's how I'd notice, all of sudden all of my services would error out because they couldn't write to disk. On all my servers /vms1 was a raid 5 on a proper raid controller showing no errors. I rebuilt one host machine using lvm + dmraid to try something new and that hasn't had any errors yet (but it has only been a few days).
To make matters worse, some corrupted the snapshot files and vmdk files when I tried upgrading vbox to fix the issue and all of a sudden the virtualmachines were listed as "innaccessable" on vboxmanage list vms, so on those hosts I ended up blowing away all the images / vmdk files and virtualbox config files.
comment:3 by , 13 years ago
Replying to frank:
Your host kernel is 2.6.38 which does not have this problem (it was fixed in 2.6.36). So the question is, where your corruptions come from. Sorry for asking, I have to ask this: Did you always terminate the VMs properly or did some VMs crash which could have resulted in such a corruption? And does your host disk which is mounted at /vms1 behave correctly or does it show any errors?
One question for you, I do have several issues with centos and fedora guests where when I shut down inside the virtualmachine, the disks don't completely unmount, so I sometimes have to force a poweroff. That could explain some issues with those vm's, but not ubuntu / solaris / opensolaris.
comment:4 by , 12 years ago
I think I found and fixed the problem. Can you please try this test build?
Thanks for the report.
Your host kernel is 2.6.38 which does not have this problem (it was fixed in 2.6.36). So the question is, where your corruptions come from. Sorry for asking, I have to ask this: Did you always terminate the VMs properly or did some VMs crash which could have resulted in such a corruption? And does your host disk which is mounted at /vms1 behave correctly or does it show any errors?