Hi Russel, Sorry for late response, there are different types of errors in 8.0 Fuel (nodes->error_type field in db), if error caused by provisioning, those nodes which have state=error, error_type=provision, will be re-provisioned, with possible data lose, those node which failed to deploy state=error, error_type=deploy will be only redeployed, which means Fuel just reruns `puppet apply` second time, which should be pretty safe operation to do. But don't forget to do backups and check your changes on staging (it's a good thing to do anyway).
Thanks, On Tue, Nov 8, 2016 at 4:12 PM, Russell Holloway < russell.hollo...@hotmail.com> wrote: > Hi all, > > > I'm working with Fuel 8 and am wondering if anyone can help me understand > how errors are handled in Fuel and deployment in general. I have a cluster > that I'm having issues deploying a second controller to and upon failure it > originally flagged all existing nodes + the new node as 'error' in Fuel. > > > When this happens, the Fuel UI unlocks configuration of the nodes in > 'error' state as if to let me reconfigure / adjust them and > reprovision+deploy, and I can hit deploy again on the dashboard. However, > what happens if I do hit deploy? Will it wipe my existing nodes and > rebuild? That would not be ideal, especially for storage nodes. Or, will it > re-run puppet scripts against the nodes without removing anything such as > data, just copying over configs and making sure they are in line with what > Fuel has / restart services? > > > I wasn't sure, so I modified the states manually in Fuel database for > cluster to operational, existing nodes to ready, and new node to > discover (which probably isn't ideal either but hopefully I didn't break > anything terribly). The only side affect I've noticed is the new node > (node-10) is showing up in haproxy configuration on the existing controller > even though it was removed from environment. There are probably other > artifacts that may be whack. > > > In any case, mostly curious how errors are handled and if Fuel has a way > to re-deploy configs and existing services as needed to an existing cluster > (either in ready state or error state) without causing data loss by wiping > storage (ceph OSDs in my case) or other things that would not be great to > lose. > > > -Russ Holloway > > > > > _______________________________________________ > Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/ > openstack > Post to : openstack@lists.openstack.org > Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/ > openstack > >
_______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : openstack@lists.openstack.org Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
