Re: [Xen-users] repeated DomU volume corruption

[Robert Rust <rjrbytes@xxxxxxxxx>]

I thought I'd provide an update on this. After fighting with this for a while (nobarrier didn't seem to eliminate the problem), I decided to put my experimenting with Xen on hold for a while. I'm now running KVM (using Proxmox), which hasn't exhibited any issues in the 2-1/2 weeks I've had it up (7 VMs). I don't know if the corruption was a Xen issue or a kernel issue or maybe a driver issue, but I need to be able to trust my system, even if it means a hit with performance. If I end up with a spare hard drive down the road, I could do some more experimenting, but for now I'll just keep an eye on the Xen project while running kvm. If anybody wants further details on my setup (e.g. hardware, specific OS version, etc.), I'd be happy to offer up what I can.

-Robert

On Mon, Dec 8, 2014 at 8:04 AM, Robert Rust <rjrbytes@xxxxxxxxx> wrote:

Peter -

It looks to me like barriers were disabled:

[ Â 28.773243] EXT4-fs (xvda2): re-mounted. Opts: nobarrier,errors=remount-ro

As far as the IO errors, I've not seen indications of it on the console, but I can verify with more certainty the next time it happens. Are there log files that would be useful to check at the Dom0 level?

-Robert

On Mon, Dec 8, 2014 at 4:48 AM, Peter Viskup <skupko.sk@xxxxxxxxx> wrote:

I was experiencing similar issue on CPU and IO bound domU and barrier=0/nobarrier helped me.

Check whether the barriers were disabled (dmesg| awk '/xvda/||/barrier/').

Check whether some IO errors are visible on Dom0 as it provides the storage layers.

--

Peter Viskup

On Sat, Dec 6, 2014 at 1:27 PM, Robert Rust <rjrbytes@xxxxxxxxx> wrote:

I'm not sure there's been a significant improvement. I just did package updates and 4 of my 7 VMs ended up exhibiting this problem. One may have already been in that state when I started (hard to tell), a second appeared to have it happen mid-shutdown, a third had it happen after it completed downloading the package updates but hadn't installed them yet, and the 4th I'm not certain of. All 4 had the nobarrier option in their fstab:

/dev/xvda2 / ext4 nobarrier,noatime,nodiratime,errors=remount-ro 0 1

Further thoughts?

-Robert

On Mon, Nov 24, 2014 at 8:29 PM, Robert Rust <rjrbytes@xxxxxxxxx> wrote:

On Mon, Nov 24, 2014 at 1:10 PM, Steve Dawson <sdawson@xxxxxxxxxxxxxxxx> wrote:

I had an issue similar in nature, which manifested as a missing partition table after a domU reboot, mounting the xvda{x} partitions with the "nobarrier" mount option resolved it. The issue was observed on xfs and ext4 filesystems.

Hope this helps.

Steve

Okay. I've added the nobarrier option to all my DomUs ... we'll see how that goes. Here's hoping it's that simple ...

-Robert

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users