Open Stack

On Mon, 2021-08-30 at 18:31 +0200, Ignazio Cassano wrote:
> Hello Sean, I am agree to use a testing environment before running on
> production.
> I wonder if kolla detects the presence of new images and makes all needed
> steps for stopping containers and restarting from new images.
if you are not changing the tag then no kolla wont see that the tag was updated on docker hub and pull the new images

if you are building your own images and pushign them to a local registry with a new tag e.g. train-30-08-2021 and then you
update the tag in your gloabl.yaml kolla will detech that that image is not avaiable local on the nodes, pull it and then update it.

generally i would suggest when you pull the images to your local regestry you tag them your self with a new tag and then update
the tag in your gloabl.yml so that you can roll back to the old images eaislly if you need too.


> I am new on docker.
> I am asking because in my last update I got same issues in hacluster and
> rabbitmq.
> I solved stopping, remove related containers and volumes and deploying
> again.
> Ignazio
> 
> 
> Il Lun 30 Ago 2021, 18:19 Sean Mooney <smooney at redhat.com> ha scritto:
> 
> > On Mon, 2021-08-30 at 17:46 +0200, Ignazio Cassano wrote:
> > > Hello, I would like to ask if the procedure I am using for update my
> > kolla
> > > wallaby is correct.
> > > Every week new images are released so I execute the following steps:
> > > 
> > > 1. Pull new images on my local registry
> > > 
> > > 2 pull new images from my local registry to controllers and compute nodes
> > > 
> > > 3 kolla-ansible deploy
> > > 
> > > Are the above steps the correct way or I missed something?
> > > Must I clean something before deploying ?
> > > Thanks
> > 
> > that is more or less what i have seen recommended in the past yes.
> > 
> > you cna use deploy or reconfigure.
> > in generall kolla-ansible upgrade is not needed unless you are doing a
> > major version bump
> > so in your case pulling the images then running deploy should simply
> > replace the containers.
> > 
> > one thing you are proably missing is removing unused un tagged images form
> > the nodes.
> > e.g. you will want to perodically remove the old images so that they disks
> > dont fill up.
> > 
> > you also proably want to add a staging step between 1 and 2 wheere you
> > test thsi on a small test env first just to make sure there
> > are no surpriese before pushing to production.
> > > 
> > > Ignazio
> > 
> > 
> >