From mboxrd@z Thu Jan 1 00:00:00 1970 From: Loic Dachary Subject: Re: Teuthology Date: Mon, 28 Mar 2016 13:17:00 +0200 Message-ID: <56F912AC.40202@dachary.org> References: <56F00270.5010901@dachary.org> <56F900EE.90106@dachary.org> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from relay6-d.mail.gandi.net ([217.70.183.198]:50412 "EHLO relay6-d.mail.gandi.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751444AbcC1LRH (ORCPT ); Mon, 28 Mar 2016 07:17:07 -0400 In-Reply-To: Sender: ceph-devel-owner@vger.kernel.org List-ID: To: Herman Narkaytis Cc: Vitalii Solovei , Ceph Development , Maksym Shalamov , Thierry Delamare Hi Herman, I recently succeeded (thanks to Thierry's help ;-) to install paddles u= sing ansible[1]. The playbook was developped recently by Zack. Even if = you don't otherwise use ansible, it's a better option than more complex= instructions. And easier to maintain as you can rely on Andrew, Dan an= d Zack (and maybe others) to maintain the playbook. One thing that I li= ke (but is frowned upon by more seasoned ansible users) is to use ansib= le from the node where paddles is to be installed. For instance: ansible-playbook -i paddles, -c local paddles.yml I suppose the http://docs.ceph.com/teuthology/docs/LAB_SETUP.html docum= ent should be updated with this new recommended deployment method (the = main document is at https://github.com/ceph/teuthology/blob/master/docs= /LAB_SETUP.rst). Toward the end of your document, the "Submitting slave node" is where t= he I don't have any experience. But David Galloway and Dan Mick and may= be others do as they have deployed and configured all the hardware curr= ently in the sepia lab. It's a significant undertaking but is fortunate= ly well documented and all files you need are public. I chose an easier alternative which is to rely on an OpenStack cluster = to provision the resources required to run teuthology jobs[2]. The last pitfall that comes to mind, and it's a significant one, is tha= t a number of teuthology jobs are sensitive to the infrastructure. For = instance, I recently worked around a paramiko bug that only showed on O= penStack although it is by no mean specific to OpenStack[3]. Fixing cep= h-qa-suite so that it does not break because the infrastructure is slow= er or faster, has 1G network or 10G, has fast or slow internet connecti= ons etc. is sometime non trivial. But for the most part it's easy and a= little tedious: you just have to do it on a regular basis :-) My 2cts [1] installing paddles via ansible http://www.spinics.net/lists/ceph-de= vel/msg28974.html [2] OpenStack backend https://github.com/dachary/teuthology/tree/openst= ack#openstack-backend [3] openstack: SSHException: Key-exchange timed out http://tracker.ceph= =2Ecom/issues/15236 On 28/03/2016 12:14, Herman Narkaytis wrote: > We intended to publish this instruction as public document, so feedba= ck from broader community would be valuable. >=20 > HHN. >=20 > On Mon, Mar 28, 2016 at 1:01 PM, Loic Dachary > wrote: >=20 > Hi, >=20 > On 28/03/2016 10:55, Herman Narkaytis wrote: > > Loic, > > Below you will find a link on Teuthology deployment instructi= ons. I would be appreciated to get your feedback on it. >=20 > This is a precise and detailed documentation, very useful. Do you= mind if I post my remarks about it with ceph-devel in cc ? Other devel= opers will certainly have interesting contributions to make. >=20 > > https://docs.google.com/document/d/18MW0NAwEZB45lGgsgjpb4fFdi= 8ope1YGr5OA-uvsIR8/edit > > > > We are still figuring out how to deploy GitBuilder, but it sh= ould not be as hard as it was with Teuthology. > > HHN. >=20 > :-) >=20 > Cheers >=20 --=20 Lo=C3=AFc Dachary, Artisan Logiciel Libre -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" i= n the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html