[ovs-discuss] Help determining the cause of a kernel panic...

[Travis Wilson]

Okay, good news. I was able to get my br-tun interfaces back online and can
confirm that we are still not crashing. I still can't get my instances to
get IPs yet, but working to get that fixed next. Thank you.


On Thu, Jan 23, 2014 at 9:27 AM, Travis Wilson <cloudtester1225 at gmail.com>wrote:

> Okay, I may have spoken too soon. On further investigation of my instances
> not working issue I discovered this morning that after my upgrades
> yesterday my br-tun instances had been removed from both my network and
> compute nodes. I suspect that is why 1.9.3 and 2.0.0 didn't crash initially
> for me. I'm attempting to get those back online today. Once I have them
> recreated I'll report back on whether the new versions actually solved my
> issues. My apologies, I totally missed the fact that they were missing
> initially.
>
> Thank you.
>
>
> On Wed, Jan 22, 2014 at 3:01 PM, Jesse Gross <jesse at nicira.com> wrote:
>
>> Hmm, that's interesting. I'm glad that 1.9.3 fixed the problem
>> although I didn't see any changes from 1.9 that I would expect to
>> resolve this. However, 2.0 has a much simpler tunneling stack so I
>> would expect that to be more reliable.
>>
>> On Wed, Jan 22, 2014 at 11:25 AM, Travis Wilson
>> <cloudtester1225 at gmail.com> wrote:
>> > Good day again Jesse.
>> >
>> > I wanted to provide you and the group with an update on our issue. We
>> tried
>> > several times to get a kdump out of the system but never were able to
>> > successfully get that to work. Based on that we decided to take a look
>> at
>> > the newer versions of Open vSwitch as you had mentioned that those might
>> > improve the issue we were seeing. Following this guide
>> > (
>> https://github.com/mininet/mininet/wiki/Installing-new-version-of-Open-vSwitch
>> )
>> > I initially downloaded and installed version 1.93 of Open vSwitch. I
>> don't
>> > know what it fixed, but as soon as that version was loaded I was able to
>> > start instances without the network or the compute nodes crashing. I so
>> > still have other deployment issues, but at least the vSwitch issue
>> seemed to
>> > be resolved.
>> >
>> > Since 1.9.3 was working I decided to just go ahead and upgrade both of
>> the
>> > nodes running Open vSwitch to 2.0.0 so that we are running with the
>> latest
>> > and greatest edition on them. So far no additional crashes have
>> occurred.
>> >
>> > As soon as time allows I will load the 1.9.0 edition on my home system
>> and
>> > reproduce the original issue I saw and see if I can get a kdump from
>> it, but
>> > for now the newer editions have solved my immediate issues.
>> >
>> > Thank you again for all of your time investigating and assisting me with
>> > these issues. It is appreciated.
>> >
>> > Have a great week!
>> >
>> >
>> > On Fri, Jan 17, 2014 at 12:09 PM, Jesse Gross <jesse at nicira.com> wrote:
>> >>
>> >> Thanks, this is useful.
>> >>
>> >> If you have the binary and can use gdb to find what line is causing the
>> >> problem based on the instruction pointer, that would be very helpful. I
>> >> agree that it seems unlikely that this is memory corruption.
>> >>
>> >> On Fri, Jan 17, 2014 at 5:51 AM, Travis Wilson <
>> cloudtester1225 at gmail.com>
>> >> wrote:
>> >>>
>> >>> Good day.
>> >>>
>> >>> So I am working with one of our internal Linux resources trying to get
>> >>> kdump to capture a proper dump. So far we haven't had any luck with
>> getting
>> >>> kdump to capture things properly. I did however get a bit of a
>> different
>> >>> screenshot of a crash today and wanted to share it in case it would be
>> >>> helpful.
>> >>>
>> >>> Here is what we saw today when the crash occurred.
>> >
>> >
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20140123/aa38cc09/attachment.html>