Re: [Xen-devel] Unstable NFS mount at heavy load.

["G.R." <firemeteor@xxxxxxxxxxxxxxxxxxxxx>]

On Wed, Jan 23, 2013 at 4:29 AM, Konrad Rzeszutek Wilk
<konrad.wilk@xxxxxxxxxx> wrote:
> On Mon, Jan 21, 2013 at 12:01:43AM +0800, G.R. wrote:
>> On Sat, Jan 19, 2013 at 12:14 AM, Konrad Rzeszutek Wilk
>> <konrad.wilk@xxxxxxxxxx> wrote:
>> > On Wed, Jan 16, 2013 at 12:50:08AM +0800, G.R. wrote:
>> >> Hi Konrad, do you have any suggestion how to debug?
>> >
>> > Is your dom0 32-bit or 64-bit? And what kind of network card are you
>> > using for the NFS traffic?
>> >
>> I have both 64-bit dom0 && domU.
>> The physical card I have is RTL8111/8168B (rev06) (10ec: 8168).
>> And the virtual card I used is e1000, but I guess this is not
>> important since I've seen this in the log:
>> Jan  6 01:31:03 debvm kernel: [    0.000000] Netfront and the Xen
>> platform PCI driver have been compiled for this kernel: unplug
>> emulated NICs.
>>
>> I'm thinking of dumping the traffic to check when I got spare time.
>> Do you think this is a good idea or do you have other suggestion?
>
> Well, the thread on "Fatal crash on xen4.2 HVM + qemu-xen dm + NFS"
> seems to imply that this a problem with NFS tcp-retransmit.
>
> And I've seen similar issues as well - but only on skge, tg3, and
> r8169 - but only when using the 32-bit domain0.
> I don't know if the issue I am hitting is the same thing.
>

I checked the thread and unfortunately did not find anything conclusive.
In my case, my dom0 seems to work fine and even the domU is still alive
-- everything back to order after the mount recovered (typically in a
couple of minutes).

According to the traffic I captured, server is kind of busy and keep
sending ZeroWindow for a while
and the client in domU reset the connection after retrying 6 times
within 15 seconds.
I'm not sure if this is a correct client behavior while the server is
doing wrong.
But why does this only happen in domU client?

Please find the traffic log in the attached file.
I've captured the traffic from both server and domU.
And it appears that there is no mismatch.

Thanks,
Timothy

Attachment: server.view
Description: Binary data

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