Actually I completely agree with you regarding doing it in devstack 1st.
However, the next step will be doing one's own deployment via puppet or
similar and creating guest images. Which gets us back to the original
issues at the start of the thread.
regards
Mark
On 08/03/17 03:33, Amrith Kumar wrote:
Yes, the documentation cannot (unfortunately) cover each and every possible
deployment scenario that one finds out there. In the case of Trove where
guest images contain software like databases, the project cannot legally
distribute guest images.
Where we provide tools to build images then those tools will be with
configurations that we can know and control; hence the project formerly
known as trove-integration (dead and gone, may it rest in peace) which has
now been integrated into the trove repository as the trovestack tool. But
that tool can only generate images which we know will work in a system
configured in a certain way. You can run the trovestack build-image command
and make just the images or you can use the kick-start mechanism which will
get you going in the full devstack environment. Whether an image built using
the build-image command will work on your specific OpenStack, one cannot say
for sure. That is just a fact that people like to ignore.
I'm always criticized for telling people to first try and deploy trove using
devstack and no one wants to listen to the simple reason for this. Trove has
a number of moving parts and a specific aspect of complexity that isn't
found in other projects; the guest agent on the guest image. And the fact
that it is a consumer of a number of underlying OpenStack services that are
each tweaked just a little bit by each deployer.
The benefit of kicking off Trove in a simple VM with devstack and using the
devstack tools to get things going is that you see how the whole system is
supposed to work together and it is then much easier to go from there to
building your own images and deploying them on an OpenStack system
configured in your own specific way. And yes, every single OpenStack system
that I've seen is configured in some slightly different way and many of
these apparently inconsequential changes do in fact impact Trove.
I'll offer the same suggestion again, get a simple system up and running
with devstack and see how the piece parts fit together in a configuration
that is supposed to work. That'll make it easier to understand how things
are supposed to work in whatever configuration you want to put together.
This was the same advice I was given when I started working on Trove, and
believe me, it is good advice.
And if the documentation is lacking, and you find something helpful (or
unhelpful) in the documentation, maybe contributing it would be useful for
others.
-amrith
-----Original Message-----
From: Mark Kirkwood [mailto:mark.kirkw...@catalyst.net.nz]
Sent: Tuesday, March 7, 2017 12:15 AM
To: openstack@lists.openstack.org
Subject: Re: [Openstack] Trove is skipped in Newton
I haven't tried to get Trove working with latest versions - but experience
with
previous (Kilo, Liberty, Mitaka) ones suggests that the creation of guest
images requires painstaking determination and error checking - as the
supplied docs are - somewhat - lacking...(read between the lines here).
Best wishes
Mark
On 07/03/17 20:13, Evgeniy Ivanov wrote:
Hello everybody,
I found some interesting info about Trove in Newton. I'm able to
deploy successfully some trove components (api, conductor and
taskmanager) on my test Newton environment. As for the Trove
guestagent - I'm totally not able to run it neither on Ubuntu Trusty
nor Ubuntu Xenial.
I found there are no Newton packages for Trusty in The Ubuntu Cloud
Archive. The last supported release is Mitaka (I checked the
RELEASE_MAP). That looks there is no support for Newton in Trusty,
therefore I can not use Trove in Newton.
As for the Xenial release, here is a bug where Amrith Kumar (CTO
Tesora) says there is no support for Newton also -
https://bugs.launchpad.net/trove/+bug/1652312 (Should be in Ocata
only)
For me, it looks like Tesora just skipped Newton, did they?
I'm asking here because there is no way to get any answers from Tesora
and Co.
Thanks for any help!
_______________________________________________
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
_______________________________________________
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
