*** rpittau|afk is now known as rpittau | 06:22 | |
opendevreview | Michal Arbet proposed openstack/kolla master: Allow build debian/ubuntu-binary cloudkitty images https://review.opendev.org/c/openstack/kolla/+/797433 | 06:40 |
---|---|---|
opendevreview | Michal Arbet proposed openstack/kolla master: ironic-neutron-agent: enable for Debuntu binary https://review.opendev.org/c/openstack/kolla/+/799611 | 07:00 |
opendevreview | Michal Arbet proposed openstack/kolla master: Allow build debian/ubuntu-binary cloudkitty images https://review.opendev.org/c/openstack/kolla/+/797433 | 07:03 |
kevko | \o/ | 07:05 |
mgoddard | morning | 07:45 |
mgoddard | yoctozepto: I think your framework is actually a toolkit | 07:45 |
mgoddard | https://martinfowler.com/bliki/InversionOfControl.html | 07:46 |
opendevreview | Mark Goddard proposed openstack/kayobe-config-dev stable/wallaby: Support specifying a Kayobe environment to use https://review.opendev.org/c/openstack/kayobe-config-dev/+/804332 | 07:56 |
opendevreview | Mark Goddard proposed openstack/kayobe-config-dev stable/wallaby: Support reading Kayobe environment name from file https://review.opendev.org/c/openstack/kayobe-config-dev/+/804333 | 07:56 |
opendevreview | Mark Goddard proposed openstack/kolla-ansible stable/wallaby: fluentd: Fix check for external elasticsearch https://review.opendev.org/c/openstack/kolla-ansible/+/804334 | 07:57 |
kevko | hello guys .. | 07:59 |
opendevreview | Mark Goddard proposed openstack/kolla-ansible stable/victoria: fluentd: Fix check for external elasticsearch https://review.opendev.org/c/openstack/kolla-ansible/+/804357 | 08:00 |
kevko | i was just testing upgrade nova (compute part - two computes) and I wonder why from 5 instances 2 instances are shutdown (compute0,compute1) while other instances are normally alive ? I know that it's no kolla related ...but have you seen anyone before ? | 08:01 |
kevko | normally i am migrating all to another compute ..but i was curious if it is OK to upgrade online | 08:01 |
opendevreview | Pierre Riteau proposed openstack/kolla master: [release] Use RDO master Delorean packages https://review.opendev.org/c/openstack/kolla/+/804269 | 08:07 |
priteau | mgoddard: I thought of a simpler approach for the neutron api-paste kolla/kolla-ansible dependency | 08:08 |
priteau | I can switch to xena and keep using the old file location for now | 08:08 |
opendevreview | Pierre Riteau proposed openstack/kolla-ansible master: Use api-paste.ini from /etc/neutron https://review.opendev.org/c/openstack/kolla-ansible/+/804288 | 08:11 |
priteau | Sorry for the double review | 08:13 |
yoctozepto | mgoddard: well, the project's certainly both as it is meant to deliver the tools as well as the framework (the framework "frames" [defines] the way of thinking or working with it) | 08:14 |
mgoddard | the code it produced would not be a framework | 08:14 |
yoctozepto | mgoddard: the way to structure the roles would be | 08:15 |
mgoddard | it's almost the opposite | 08:15 |
mgoddard | if we allowed users to replace the service* roles with their own, we could call kolla-ansible a framework | 08:16 |
mgoddard | but we won't do that | 08:16 |
yoctozepto | I think you are drawing a wrong boundary; the boundary is on project roles; the service roles compose the accompanying toolkit | 08:17 |
mgoddard | the project roles are not a framework, since the service roles are not intended to be replaceable | 08:19 |
yoctozepto | moreover, this framework is shifting towards accepting more declarative than imperative code | 08:19 |
mgoddard | agreed, service roles are a toolkit | 08:19 |
yoctozepto | yes, that we agree on :-) | 08:19 |
yoctozepto | the way we enforce the structure of the project roles is the framework | 08:20 |
yoctozepto | https://en.wikipedia.org/wiki/Software_framework | 08:20 |
mgoddard | as a user of a framework, I do not modify the framework code. I plug my code into it, and it executes my code at the right time | 08:20 |
mgoddard | there is no other code in our case, only config | 08:21 |
mgoddard | we're just building an application (kolla-ansible) that uses a toolkit (service-*), and accepts config | 08:21 |
yoctozepto | the config is code; the code is config | 08:22 |
mgoddard | sort of, but I draw a distinction between globals.yml and a playbook/role | 08:22 |
mgoddard | especially since users modify globals.yml, but not playbooks or roles | 08:23 |
yoctozepto | yes, but notice the framework is not for users of kolla-ansible but for us, cores, and contributors, and extenders, and general tinkerers | 08:24 |
yoctozepto | it is about maintainability and extensibility with DRY | 08:24 |
yoctozepto | and I assure you it matches the definition of a framework | 08:24 |
mgoddard | I disagree | 08:25 |
mgoddard | despite your assurances | 08:25 |
mgoddard | perhaps you have some additional magic planned | 08:27 |
mgoddard | if you can demonstrate inversion of control, I may come around | 08:28 |
yoctozepto | the way we have dedicated playbooks per kolla ansible action is actually a way of ioc possible in ansible | 08:33 |
yoctozepto | this is already a part of the "latent framework" that I discuss there | 08:33 |
yoctozepto | do note we are not dealing with a general purpose programming language exercise here so comparing with, e.g., web frameworks is going to be blurry ;-) | 08:35 |
mgoddard | here's how I see a framework working in ansible | 08:41 |
mgoddard | it includes a playbook | 08:41 |
mgoddard | and includes a role with a dynamic name | 08:41 |
mgoddard | as a user of the framework, you provide a role name via a variable | 08:41 |
mgoddard | the role must accept specific variable names | 08:42 |
mgoddard | the playbook might setup a DB, LB and app server | 08:42 |
mgoddard | and your role deploys your code | 08:43 |
mgoddard | obviously a basic example, but the key point is that the framework's playbook is in control, and calls out to code provided by the developer | 08:43 |
mgoddard | anyway, enough debate. There's time to change my mind | 08:46 |
yoctozepto | yeah, we call out to provided roles using chosen playbook as an entry point, the role should consume certain variables - all in check - the latent framework is already here :-) | 08:48 |
mgoddard | disagree | 08:49 |
mgoddard | I cannot add a new role without modifying the playbook | 08:49 |
yoctozepto | yes, because this one part is rigid now | 08:51 |
yoctozepto | we can agree that the framework is handicapped for now | 08:51 |
mgoddard | it's just a pattern | 08:51 |
yoctozepto | ... that you must follow to get it right | 08:52 |
yoctozepto | you can abuse web frameworks as well | 08:52 |
yoctozepto | some assumptions are only in docs | 08:53 |
yoctozepto | not verifiable by the framework | 08:53 |
yoctozepto | it's just how life goes :-) | 08:53 |
yoctozepto | in ansible it's even worse | 08:53 |
yoctozepto | ok, enough debate really; whatever we call it's meant to increase the productivity and not start flame wars :-) | 08:53 |
jingvar | hi! can someone help to undertand how to run node config generator ("fake") | 09:02 |
mnasiadka | I love academic debates in the morning ;) | 09:02 |
kevko | finally there is some discussion to read :D | 09:10 |
opendevreview | Michal Arbet proposed openstack/kolla master: Allow build debian/ubuntu-binary cloudkitty images https://review.opendev.org/c/openstack/kolla/+/797433 | 09:13 |
kevko | ./tools/kolla-ansible -i /etc/kolla/inventory upgrade -t neutron --limit compute <<< should this run following tasks ? | 09:14 |
kevko | Running Neutron database expand container | 09:14 |
kevko | Checking neutron pending contract scripts | 09:14 |
kevko | Stopping all neutron-server for contract db | 09:14 |
kevko | etc.. ? | 09:14 |
kevko | because firstly it is failing ..and secondly ..i think it should not run any db migrations if limiting to compute group, shouln't it ? | 09:15 |
yoctozepto | hmm, indeed it seems weird to match these tasks on computes | 09:17 |
opendevreview | Merged openstack/kolla master: Fix naming of CentOS Stream in reno https://review.opendev.org/c/openstack/kolla/+/804251 | 09:17 |
yoctozepto | consider it a bug and fix | 09:17 |
opendevreview | Radosław Piliszek proposed openstack/kolla stable/wallaby: Fix naming of CentOS Stream in reno https://review.opendev.org/c/openstack/kolla/+/804336 | 09:18 |
opendevreview | Radosław Piliszek proposed openstack/kolla-ansible stable/ussuri: fluentd: Fix check for external elasticsearch https://review.opendev.org/c/openstack/kolla-ansible/+/804337 | 09:20 |
mgoddard | it's due to run_once | 09:23 |
mgoddard | ideal solution would be a separate play/role, similar to nova/nova-cell | 09:24 |
mgoddard | but that's a bigger job | 09:24 |
yoctozepto | we love ansible quirks | 09:33 |
mnasiadka | does it mean we should split all roles into -server and -agent side? :) | 09:33 |
yoctozepto | would not hurt :-) | 09:34 |
yoctozepto | ping cores on https://review.opendev.org/c/openstack/kolla-ansible/+/787875 - feedback welcome | 09:44 |
kevko | mnasiadka: so you want to generate new bugs :) ? | 09:45 |
mnasiadka | kevko: if we decide it's boring enough to do so - why not ;) | 09:45 |
mgoddard | most roles don't have a compute component, but those that do might be good to split | 09:56 |
mgoddard | at least from a performance perspective | 09:57 |
yoctozepto | indeed, due to how ansible works | 09:59 |
yoctozepto | in fact, it makes sense because the *role* to be a compute target is different from being a controller | 10:00 |
yoctozepto | our default split is service-centric instead of node-centric | 10:00 |
yoctozepto | and this causes hiccups | 10:00 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: Validate if running CentOS OS is CentOS Stream https://review.opendev.org/c/openstack/kolla-ansible/+/804072 | 10:08 |
opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Trivial fix nova's healthchecks https://review.opendev.org/c/openstack/kolla-ansible/+/804381 | 11:18 |
kevko | ^^ please approve this after zuul +1 as this is very trivial fix | 11:19 |
yoctozepto | kevko: commented | 11:38 |
kevko | yoctozepto: aaaa, i forgot | 11:39 |
kevko | done | 11:40 |
opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Trivial fix nova's healthchecks https://review.opendev.org/c/openstack/kolla-ansible/+/804381 | 11:40 |
opendevreview | Michal Nasiadka proposed openstack/kolla-ansible master: Validate if running CentOS OS is CentOS Stream https://review.opendev.org/c/openstack/kolla-ansible/+/804072 | 12:03 |
opendevreview | Radosław Piliszek proposed openstack/kolla-ansible master: Validate if running CentOS OS is CentOS Stream https://review.opendev.org/c/openstack/kolla-ansible/+/804072 | 12:45 |
opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Fix neutron upgrade when running with limit https://review.opendev.org/c/openstack/kolla-ansible/+/804384 | 13:24 |
opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Fix neutron upgrade when running with limit https://review.opendev.org/c/openstack/kolla-ansible/+/804384 | 13:26 |
yoctozepto | mgoddard: not triggered by Ironic being called a framework? :-) | 14:04 |
mgoddard | in this case, not so much! | 14:04 |
yoctozepto | but still a bit I see! haha | 14:05 |
mgoddard | the driver model is closer to a framework | 14:05 |
opendevreview | Michal Arbet proposed openstack/kolla-ansible master: Use Docker healthchecks for nova-spicehtml5proxy service https://review.opendev.org/c/openstack/kolla-ansible/+/804391 | 14:20 |
kevko | yoctozepto: what do you mean because of ansible quirk ? | 14:22 |
kevko | yoctozepto: i was testing it and it's working ..but not saying it is wrong approach from my side | 14:22 |
jingvar | I see Mark :) | 14:26 |
yoctozepto | kevko: https://github.com/ansible/ansible/issues/11496 | 14:28 |
yoctozepto | mgoddard can confirm | 14:29 |
kevko | ok, i will rework then | 14:30 |
kevko | but i've sent you already another trivial patch :P | 14:30 |
opendevreview | Pierre Riteau proposed openstack/kayobe master: Fix documentation links for routed control plane networks https://review.opendev.org/c/openstack/kayobe/+/804398 | 14:52 |
opendevreview | Pierre Riteau proposed openstack/kayobe master: Fix documentation links for routed control plane networks https://review.opendev.org/c/openstack/kayobe/+/804398 | 15:03 |
*** rpittau is now known as rpittau|afk | 15:31 | |
opendevreview | Merged openstack/kolla stable/wallaby: Fix naming of CentOS Stream in reno https://review.opendev.org/c/openstack/kolla/+/804336 | 15:38 |
yoctozepto | heads-up: we are going to retire murano (cc mgoddard who had this wish) | 15:55 |
yoctozepto | (continue wishing) | 15:55 |
opendevreview | Radosław Piliszek proposed openstack/kolla-ansible master: Bump libvirtd memlock ulimit https://review.opendev.org/c/openstack/kolla-ansible/+/804450 | 18:39 |
opendevreview | Radosław Piliszek proposed openstack/kolla-ansible master: [CI] Fix Debian upgrade https://review.opendev.org/c/openstack/kolla-ansible/+/804455 | 19:15 |
opendevreview | Radosław Piliszek proposed openstack/kolla-ansible master: Detect outward-rabbitmq and rabbitmq group overlaps https://review.opendev.org/c/openstack/kolla-ansible/+/801747 | 19:20 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!