I'm at a loss for how you ended up there. I don't see any problems going
forward except that you would probably want to reboot any discover nodes on
that are not provisioned so that they are connected to the right
mcollective instance, (they won't take commands if they are pointing to the
wrong one), otherwise you are probably fine
On Mon, Sep 19, 2016 at 11:55 AM Taseer Ahmed <tasee...@gmail.com> wrote:
> I was running fuel in a VM named *fuel*. So actually what happened was
> that, while removing some of my routes from the iptables, I mistakenly
> deleted the ipaddress on eth0 (PXE network). So I lost connection with the
> fuel node. However, quite interestingly, on my base machine running the
> fuel, I was able to observe that another instance with name *fuel-master *had
> sprung up.
> It had the same IP address on eth0 as the previous one, but it did not
> have routes to the compute and controller nodes, hence the errors I
> previously asked about. For now, I have shutdown this instance, i.e
> *and rebooted the original one (fuel). Its working fine for now. It is
> able to detect the pool of unallocated machines and have also successfully
> deployed a new cluster.
> Is there any explanation for the above, and would it cause complexities in
> the future ?
> Thank you.
> On Wed, Sep 14, 2016 at 8:20 PM, Andrew Woodward <xar...@gmail.com> wrote:
>> Given it strongly sounds like somehow we managed to re-install the fuel
>> node, you should not reboot the nodes without taking care to disable boot
>> from network.
>> The way we handle provisioning, we leave an entry in cobbler to control
>> the bootp profile so that it boots from local disk in the event its a
>> provisioned node, given that these entries would not exist on a fresh
>> install, you would find that the nodes will boot into the discover image as
>> if they where un-provisioned nodes rather back to the installed os.
>> On Wed, Sep 14, 2016 at 5:48 AM Taseer Ahmed <tasee...@gmail.com> wrote:
>>> Thank you for your reply. I carried out the 'mco ping'. I got only one
>>> response, it was from Master. However, a simple ping generates a positive
>>> response from all the nodes in the cluster.
>>> In addition, I also checked my Fuel Master web UI. There, I found out
>>> that both, controller and compute nodes were offline. However, I can access
>>> the Horizon Dashboard on the public network. Should I manually reboot them
>>> ? Is there any way to bring them online remotely ?
>>> Thank you.
>> Andrew Woodward
opnfv-users mailing list