fuel-dev team mailing list archive

[Evgeniy L <eli@xxxxxxxxxxxx>]

Hi,

I would like to provide status for master node upgrade feature.
We merged a patch [1] which increased stability of upgrade.

It fixes a lot of problems here is the list of some of them:
* fixed known and unknown raise conditions, e.g. keystone
  db migration interruption
* now we won't have problem with ip duplication, I haven't
  seen the problem, so I cannot say how often it happened,
  but can say that the patch solves the problem in case of upgrade
* the patch twice reduces probability of docker's death during
  the upgrade

In the last 24 hours I tested the patch on 4 virtual machines,
there were *915* upgrade runs (to reduce the time of upgrade
I enabled only docker engine, which is the most problematic
part of master node upgrade), at the end of the day there
were *0* failed upgrades.

Thanks,

[1] https://review.openstack.org/#/c/118387/


On Wed, Aug 27, 2014 at 4:05 PM, Matthew Mosesohn <mmosesohn@xxxxxxxxxxxx>
wrote:

> Regarding paste.openstack.org, we should look to another pastebin
> provider. It has been giving 500 errors quite nearly consistently for
> me lately and really interferes with my work. We could use pastie.org,
> for example.
>
> On Wed, Aug 27, 2014 at 12:53 PM, Mike Scherbakov
> <mscherbakov@xxxxxxxxxxxx> wrote:
> > OpenStack uses paste.openstack.org all the time, and I've heard issues
> with
> > how long content is stored there.
> >
> >
> > On Wed, Aug 27, 2014 at 12:45 PM, Aleksandra Fedorova
> > <afedorova@xxxxxxxxxxxx> wrote:
> >>
> >>
> >> I can not find the policy about how long data is stored there, but I
> doubt
> >> that pastebin service can be used as a longterm storage. If we don't
> want to
> >> lose logs and scripts data, the use of paste.o.o links for bug reports
> >> should be forbidden.
> >>
> >> Launchpad attachments are much more reliable even though less
> comfortable
> >> to use.
> >>
> >> On Aug 27, 2014 8:49 AM, "Mike Scherbakov" <mscherbakov@xxxxxxxxxxxx>
> >> wrote:
> >>>
> >>> +1 on updating bug descriptions (not comments) about probability of
> >>> failure, and using paste.openstack.org more often.
> >>>
> >>>
> >>> On Wed, Aug 27, 2014 at 3:41 AM, Dmitry Borodaenko
> >>> <dborodaenko@xxxxxxxxxxxx> wrote:
> >>>>
> >>>> On Tue, Aug 26, 2014 at 12:11 AM, Mike Scherbakov
> >>>> <mscherbakov@xxxxxxxxxxxx> wrote:
> >>>> > "confusing versioning in OpenStack patching" - if we didn't change
> >>>> > puppet
> >>>> > manifests and Fuel/OpenStack reference architecture in next Fuel
> >>>> > versions,
> >>>> > then it would be as simple as patching from 5.0 to 5.1. But it
> >>>> > appeared to
> >>>> > be more complicated system than you would initially think of, so in
> >>>> > general
> >>>> > 5.0.2 may not be equal to 5.1, that's where all things come up. If
> we
> >>>> > had
> >>>> > OpenStack upgrades, then we could just say 5.0 -> 6.0 - easy.
> >>>>
> >>>> We may have had technical reasons to make this decision, but it still
> >>>> is confusing and negatively impacts UX. I agree that having an
> >>>> incomplete feature early is better than not having it at all until
> >>>> much later, as long as we don't stop working on it until it's complete
> >>>> and these small but annoying deficiencies are addressed. Our
> >>>> experience with technical debt so far is not very reassuring.
> >>>>
> >>>> > "issues with containers" - we have same issues with everything.
> Let's
> >>>> > take
> >>>> > Galera, for example. It's just issues. We can question maturity of
> >>>> > tools we
> >>>> > use, and here I'd agree - we spent too much fixing issues around
> >>>> > Docker. At
> >>>> > the same time, if we were about taking our own journey with LXC, we
> >>>> > would
> >>>> > likely spend even more time inventing our own bicycle.
> >>>>
> >>>> You're assuming that it was just Docker as a piece of software that is
> >>>> the primary cause of all our troubles with Fuel upgrades. Docker is
> >>>> only a small part of the a much large and much more intrusive design
> >>>> decision to use containers for upgrading Fuel (and also the design
> >>>> decision to use a different mechanism based on Puppet for patching
> >>>> OpenStack). I think we should question high-level design decisions
> >>>> like these more often, even after they are implemented.
> >>>>
> >>>> > Also, I'd like to ask everyone to provide
> >>>> > such information in every bug you report if possible (or if get this
> >>>> > info
> >>>> > later, put comments): in many bug reports it is unclear to
> understand
> >>>> > how
> >>>> > severe issue is.
> >>>>
> >>>> I think we should start updating bug description more often, so that
> >>>> you can find a summary of current state of the bug and of all relevant
> >>>> information from the description, without having to scroll through
> >>>> dozens of comments. We should also use paste.openstack.org more
> >>>> heavily and avoid pasting more than 1-2 lines of logs into bug
> >>>> description and comments, also to make it easier to find important
> >>>> bits in bugs history.
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Mike Scherbakov
> >>> #mihgen
> >>>
> >>>
> >>> --
> >>> Mailing list: https://launchpad.net/~fuel-dev
> >>> Post to     : fuel-dev@xxxxxxxxxxxxxxxxxxx
> >>> Unsubscribe : https://launchpad.net/~fuel-dev
> >>> More help   : https://help.launchpad.net/ListHelp
> >>>
> >
> >
> >
> > --
> > Mike Scherbakov
> > #mihgen
> >
> >
> > --
> > Mailing list: https://launchpad.net/~fuel-dev
> > Post to     : fuel-dev@xxxxxxxxxxxxxxxxxxx
> > Unsubscribe : https://launchpad.net/~fuel-dev
> > More help   : https://help.launchpad.net/ListHelp
> >
>