| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass to callbacks https://review.openstack.org/498127 | 00:15 |
|---|---|---|
| mordred | dmsimard, jeblair: ^^ that change consumes the ara change and in local testing I get output in my log file again | 00:16 |
| mordred | dmsimard: so if I could bribe you to land that change and cut a point release, that would be stellar | 00:17 |
| mordred | the main issue is that fileConfig completely resets the entire config, so when the sqlalchemy alembic migrations get loaded after zuul_stream for the first playbook, its call to fileConfig steps on the logging config in zuul_stream | 00:19 |
| mordred | subsequent calls to ansible-playbook do not have the problem because ara only runs the migrations once | 00:19 |
| mordred | we could alternately run the alembic migrations when we do the zuul python library copy and copy a pre-built base sqlite db in to ~/.ara when we make the work dir | 00:20 |
| mordred | I like the logging config file though because it gives us a more general mechanism to do stuff should we need it | 00:21 |
| dmsimard | mordred: looking | 00:33 |
| dmsimard | mordred: added a comment, would rather we add the config option in config.py instead (and then bring it back inside env.py with current_app.config) | 00:49 |
| dmsimard | and yes, cutting a dot release is fine | 00:49 |
| dmsimard | there was some stuff I potentially needed to backport from the 1.0 branch /me looks | 00:49 |
| mordred | dmsimard: cool - and yah, config.py makes sense | 00:55 |
| mordred | dmsimard: I'll get a follow up pushed up | 00:55 |
| dmsimard | sure, I cherry-picked a patch from 1.0 but otherwise that's all I need to land. I can cut a release ... monday? :) | 00:56 |
| rcarrillocruz | dmsimard: neat, thanks for picking up | 06:47 |
| *** adam_g has quit IRC | 08:36 | |
| *** adam_g has joined #zuul | 08:41 | |
| *** adam_g has quit IRC | 08:45 | |
| *** adam_g has joined #zuul | 08:58 | |
| *** adam_g has quit IRC | 09:11 | |
| *** adam_g has joined #zuul | 09:13 | |
| *** olaph has quit IRC | 10:45 | |
| *** olaph has joined #zuul | 10:46 | |
| *** adam_g has quit IRC | 11:09 | |
| *** adam_g has joined #zuul | 11:15 | |
| dmsimard | I was thinking... https://github.com/openstack-infra/devstack-gate/blob/master/playbooks/roles/fix_etc_hosts/tasks/main.yaml feels pretty minimal | 14:32 |
| dmsimard | I would improve that to include not only the ansible_hostname fact but also the inventory_hostname (i.e, "controller") and then move that off to the v3 base job | 14:32 |
| dmsimard | Thoughts ? | 14:32 |
| jeblair | dmsimard: seems reasonable | 14:47 |
| dmsimard | ok ! | 14:47 |
| dmsimard | jeblair: found what is probably the last issue for the network overlay patch... which is related to my question about etc hosts | 14:49 |
| dmsimard | We are using inventory hostnames when setting up the bridge: http://logs.openstack.org/33/435933/40/check/gate-grenade-dsvm-neutron-dvr-multinode-ubuntu-xenial-nv/6ae0336/logs/ara/file/a096287a-1cad-4c00-8b75-d49584f41dcf/#line-11 | 14:50 |
| dmsimard | Which was probably not a bad idea but ovs doesn't like that: 2017-08-26T04:20:59.630Z|00198|socket_util|ERR|"localhost" is not a valid IP address | 14:51 |
| dmsimard | The peer inventory hostnames are their IP addresses but the primary is 'localhost' | 14:51 |
| dmsimard | I think I have a fix.. | 14:54 |
| jeblair | dmsimard: it looks like the ip was what was used before; and the last option needs the local ip regardless. is there a way to get the ip for every host, including primary? | 14:54 |
| jeblair | i've held a node to perform devstack testing | 14:55 |
| *** olaph1 has joined #zuul | 14:56 | |
| dmsimard | Yeah, but that information only lives in the nodepool subnodes file, it's hard to otherwise reliably retrieve the right IP (ex: v4 vs v6) | 14:56 |
| dmsimard | So I'm considering setting a hostvar when vm-gate-wrap creates the inventory from the subnodes file | 14:56 |
| *** olaph has quit IRC | 14:56 | |
| dmsimard | er, subnodes are already fine -- I meant /etc/nodepool/node_private | 14:59 |
| jeblair | dmsimard: maybe make it comptabile with the v3 'nodepool' hostvars? http://logs.openstack.org/99/497699/13/check/devstack-legacy-tempest-dsvm-neutron-full/2e97b8b/zuul-info/inventory.yaml | 15:00 |
| dmsimard | Good idea | 15:00 |
| jeblair | also, hrm, that private_ipv4 should not be null i don't think... | 15:01 |
| jeblair | it's set correctly in zk | 15:04 |
| dmsimard | is the role of subnodes in v3 (/etc/nodepool/role) "sub" ? (getting that in v2) | 15:06 |
| dmsimard | hm, I guess it doesn't matter. | 15:06 |
| *** olaph has joined #zuul | 15:07 | |
| *** olaph1 has quit IRC | 15:08 | |
| dmsimard | eh, sort of? there's no role hostvar right now, maybe there should be | 15:09 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul feature/zuulv3: Send all node information to executor https://review.openstack.org/498206 | 15:10 |
| jeblair | dmsimard: there's no "role" in v3 | 15:10 |
| jeblair | dmsimard: and there are no subnodes | 15:10 |
| dmsimard | ah I guess you're right | 15:11 |
| jeblair | dmsimard: they are all peers. any function that distinction serves should be handled by groups. | 15:11 |
| jeblair | dmsimard: so for the devstack-legacy job, we'll have a 'primary' node and a 'subnodes' group. | 15:11 |
| jeblair | dmsimard: but when we build the new devstack job, we'll probably start with a 'controller' node and a 'compute' group (which may include the controller node -- that's different from the current devstack-gate) | 15:13 |
| jeblair | dmsimard: so this is another instance where we may end up with a slightly different network overlay role that we use for devstack-legacy from the one that we put in the v3 multinode base job. | 15:13 |
| dmsimard | jeblair: right | 15:18 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul feature/zuulv3: Send all node information to executor https://review.openstack.org/498206 | 15:29 |
| *** olaph1 has joined #zuul | 16:03 | |
| *** olaph has quit IRC | 16:04 | |
| jeblair | mordred: https://review.openstack.org/498206 is a blocker | 16:15 |
| mordred | jeblair: looking | 16:44 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass it to callbacks https://review.openstack.org/498127 | 16:44 |
| mordred | dmsimard: ^^ remote: https://review.openstack.org/498125 Allow configuring logging via config file | 16:44 |
| mordred | dmsimard: I updated the ara patch and the zuul patch - it's now plumbed through ara's config - and is a bit nicer too | 16:45 |
| dmsimard | mordred: cool, I'll have a look after submitting my wip about inventory rework | 16:45 |
| mordred | jeblair: +A | 16:45 |
| mordred | dmsimard: cool - I'm actually pretty pleased with the new patch and think you will be too - it allows for a pretty flexible config for users without requiring it or making things more complicated | 16:46 |
| mordred | dmsimard: oh - one more thing I can do ... (did in zuul but didn't update the ara patch - it's just cosmetic, but patch coming) | 16:48 |
| openstackgerrit | Merged openstack-infra/zuul feature/zuulv3: Send all node information to executor https://review.openstack.org/498206 | 16:54 |
| dmsimard | jeblair: re-write of inventory with nodepool v3 hostvars: https://review.openstack.org/#/c/498207/ and leveraging it in network overlay patch: https://review.openstack.org/#/c/435933/40..41/playbooks/roles/ovs_vxlan_bridge_peers/tasks/main.yaml | 17:15 |
| dmsimard | it yields an inventory that looks like this (don't mind the sort-of dummy data) http://paste.openstack.org/raw/619509/ | 17:18 |
| mordred | dmsimard: so - unfortunately you do not have an inventory in place to run that create inventory playbook | 17:34 |
| dmsimard | mordred: doesn't matter, localhost is always implicitely available | 17:34 |
| dmsimard | mordred: it'll just print a warning that only localhost is available because the inventory file couldn't be found | 17:34 |
| mordred | dmsimard: ya - but you pass -i WORKSPACE/inventory to ansible-playbook | 17:34 |
| dmsimard | you need to pass an inventory file or the command doesn't run :) | 17:35 |
| dmsimard | I could pass -i foo | 17:35 |
| mordred | oh - it might be http://logs.openstack.org/07/498207/2/check/gate-tempest-dsvm-py35-ubuntu-xenial/dfda3bd/console.html#_2017-08-26_17_23_57_997240 | 17:35 |
| mordred | that's the issue | 17:35 |
| mordred | yah- log directory creation doesn't happen until 17 lines later | 17:37 |
| mordred | dmsimard: I think you can just let that one spit to stdout | 17:37 |
| dmsimard | ah, sure | 17:37 |
| dmsimard | mordred: reviewing your patch, will have a small comment otherwise lgtm | 17:38 |
| dmsimard | mordred: added a comment on your ara patch | 17:40 |
| dmsimard | I'm off for now, ttyl | 17:44 |
| mordred | dmsimard: oh great - those make me happy - fixed | 17:44 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add integration test for zuul_stream https://review.openstack.org/498209 | 18:36 |
| mordred | dmsimard, pabelanger: ^^ this is what I was thinking in terms of a simple functional test that runs ansible from one test node to a second test node with our various ansible libs installed | 18:36 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add integration test for zuul_stream https://review.openstack.org/498209 | 18:53 |
| dmsimard | mordred: makes sense, left a few comments | 19:47 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add integration test for zuul_stream https://review.openstack.org/498209 | 19:53 |
| mordred | dmsimard: thanks! fixed | 19:53 |
| dmsimard | mordred: zuul's complaining that ara is not a valid project, I guess that's because ara is not in v3 yet ? | 19:55 |
| mordred | yah - I added it to the list of things that zuulv3 knows about just a little bit ago - just need the file to get updated and zuul to notice it now | 19:56 |
| mordred | dmsimard: so i'll recheck in a little while and see how it does | 19:56 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass it to callbacks https://review.openstack.org/498127 | 20:39 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass it to callbacks https://review.openstack.org/498127 | 21:12 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass it to callbacks https://review.openstack.org/498127 | 22:02 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Write a logging config file and pass it to callbacks https://review.openstack.org/498127 | 22:31 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add unittest for secrets data https://review.openstack.org/484911 | 22:35 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Fail early if people attempt to add zuul vars or secrets https://review.openstack.org/484000 | 22:35 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Add unittest for secrets data https://review.openstack.org/484911 | 22:38 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul feature/zuulv3: Fail early if people attempt to add zuul vars or secrets https://review.openstack.org/484000 | 22:38 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!