On Mon, Apr 22, 2013 at 05:10:21PM -0700, Gurucharan Shetty wrote:
> On Mon, Apr 22, 2013 at 2:19 PM, Ben Pfaff <[email protected]> wrote:
> 
> > On Wed, Apr 17, 2013 at 03:11:12PM -0700, Gurucharan Shetty wrote:
> > > Currently, when we upgrade openvswitch packages, we do a restart
> > > of userspace daemons automatically. This does not replace the
> > > kernel module.
> > >
> > > But almost everytime, we want to use the new kernel module
> > > that comes with the new version. This means that we need to
> > > manually do a "force-reload-kmod". This step, reloads the
> > > kernel module and also restarts the userspace daemons. This gives
> > > us a total of two restarts of userspace daemons. This is quite
> > > expensive in a hypervisor with hundreds of VMs sending real traffic.
> > > This also hurts the controller as it gets two reconnections in a short
> > > amount of time.
> > >
> > > With this patch, during a package upgrade, if the kernel module
> > > on disk is different than the one that is loaded, we will
> > > automatically do a force-reload-kmod while openvswitch-switch
> > > is installed. If not, we will just do a "restart" like before.
> > >
> > > One can install the kernel package first and then install the userspace
> > > packages in 2 separate steps to enforce a single 'force-reload-kmod'.
> > >
> > > If anyone wants to just restart the userspace package instead of
> > > force-reload-kmod, they can set the value of OVS_FORCE_RELOAD_KMOD=no
> > > while installing the package.
> > > Ex: OVS_FORCE_RELOAD_KMOD=no dpkg -i openvswitch-switch*
> > >
> > > Signed-off-by: Gurucharan Shetty <[email protected]>
> >
> > This looks good.
> >
> > The one thing that it makes me wonder is whether we should print (or
> > just log) anything about what decision we're making and why.  If
> > something gets screwed up on 1 of 1000 hypervisors, then this might be
> > valuable information in the post-mortem.  (Customer: "WTF did this
> > upgrade fail?"  me: "...hmm, looks like you somehow installed a kernel
> > module a year too old." versus "...hmm, dunno.")  But it all depends
> > on how likely you think these problems are.
> >
> 
> Yes. Logging makes sense. I will send v3. I plan to use
> /sys/module/openvswitch/version
> instead of srcversion as logging that gives more meaning.

I agree that it's more meaningful.  But my assumption is that srcversion
changes whenever one modifies the module (I guess it's a hash of the
source code?), whereas version might not.  That could be a real surprise
sometimes.

Perhaps one could consider comparing and logging both version and
srcversion?
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev

Reply via email to