Is Linux 3.13, the worst kernel version, ever? YES! It is... So, why Trust come with it? WHY!?

Andy Whitcroft apw at
Mon Jul 21 19:41:29 UTC 2014

> I don't want to be rude but, Linux 3.13 sucks a lot.
> I'm wondering here, for God's sake, why Ubuntu 14.04 comes with this lame
> kernel? What Ubuntu Kernel Team have in mind?!

This kernel was selected using the standard criteria for an LTS, a kernel
which would release sufficiently early in the cycle to allow a significant
number of stable cycles to occur before release and particularly before
the first point release.  For this cycle that was v3.13.  This was
a conservative choice as this cycle is an LTS so we aim to have a
significantly longer stablisation time than that in a regular cycle.
There was no indication that this was expected to be a worse kernel
than any other we might have selected (v3.12 or v3.14).  This choice was
discussed and announced at the various UDS/vUDS/ODS sessions.

> Mark Shuttlework *must say something about this situation*. What he is
> gonna do? I really t want to know that.
> I'm a HUGE fan of Ubuntu, and I mean it, I have no plans to switch off
> Ubuntu and I'll stick with 14.04 but, definitely, Linux 3.13 was a bad, bad
> choice for a LTS version like Trusty.

[list of bugs elided for brevity]

There seem to be two separate issues triggered by QEMU/KVM setups likely
significantly exacerbated on truly NUMA kit.  From the descriptions you
posted in your original email these likely are all examples of those
two issues.

Both the Kernel and Server Teams are aware of these issues and are actively
working with the support organisation to isolate the root causes and fixes
for them and will be SRUing any such fixes as soon as sensibly possible.
In the mean time it has been recommended that KSM be disabled, and support
advisories to this effect have been issued.  Separatly we are reviewing
all of the bugs that have been reported on these kinds of issues to
categorise them and to try and confirm they are the same root cause.

> I just installed a lot of QEmu 2.0 Hypervisors with Trusty, including a
> brand new Cloud Computing with OpenStack IceHouse and, guess what?! Lots of
> problems... Because of Linux 3.13.
> *This is a unprecedented situation*, Canonical must provide a new Kernel
> version for Trusty ASAP! This is affecting hundreds of people around the
> world that are trusting in Trusty but, it is not trustable right now.   :-/
> As a workaround, I backported the Kernel from Utopic, since I need the bnx2
> kernel module (linux extra), I can not just use the mainline packages
> as-is...

As stated previously, the current recommendation is to disable KSM if you are
experiencing these issues.  Once the root cause is isolated fixes will be
released for the 3.13 based kernels via the normal updates process.

[installation instructions elided for brevity]

> IMPORTANT NOTE: I still did not tested this Linux 3.16 from Utopic on
> Trusty but, at this point, anything is better than Linux 3.13. I know that
> my environment is terrible with 3.13 and I'll report here if it becomes
> better with Linux 3.16.
> Again, Canonical, please, say something about this.
> As I said, I'm a huge fan of Ubuntu and I'm waiting to see what Canonical
> will do about this.... Canonical, your turn.

Running v3.16 is expected to work on 14.04 userspace but it is not a
recommended configuration at this time.  That said, in our testing v3.16
does not appear to trigger the same issues.  Confirmation of this on your
setup would be a useful additional data point.  Please do report any
testing of this nature back on your original bug so it can be collated
with the rest.

Be assured that people are looking at these issues as a matter of urgency.


More information about the ubuntu-server mailing list