| *** gkadam_ has joined #openstack-kayobe | 05:48 | |
| *** gkadam_ has quit IRC | 06:42 | |
| *** gkadam has joined #openstack-kayobe | 06:42 | |
| *** egonzalez_afk has joined #openstack-kayobe | 06:52 | |
| *** dougsz has joined #openstack-kayobe | 07:59 | |
| *** mgoddard has joined #openstack-kayobe | 08:03 | |
| *** raphael____ has joined #openstack-kayobe | 08:04 | |
| *** ktibi has joined #openstack-kayobe | 08:07 | |
| openstackgerrit | Mark Goddard proposed openstack/kayobe master: WIP: Use the Zuul VM as the seed https://review.openstack.org/558891 | 08:09 |
|---|---|---|
| *** raphael____ has quit IRC | 08:11 | |
| *** raphael____ has joined #openstack-kayobe | 08:15 | |
| *** raphael____ has quit IRC | 08:24 | |
| openstackgerrit | Kevin Tibi proposed openstack/kayobe master: Add support for custom repo https://review.openstack.org/559045 | 08:31 |
| ktibi | mgoddard, hi, live-restore enabled by default ? | 08:38 |
| mgoddard | ktibi: no, it's disabled by default | 08:38 |
| mgoddard | backwards compat | 08:38 |
| ktibi | ok ;) | 08:38 |
| openstackgerrit | Kevin Tibi proposed openstack/kayobe master: Support docker live-restore option https://review.openstack.org/559292 | 08:40 |
| mgoddard | ktibi: oh, you meant live-restore default in kayobe, not docker. Let's make it enabled, with an upgrade note | 08:43 |
| ktibi | yes | 08:43 |
| ktibi | hum so when you upgrade | 08:43 |
| ktibi | kayobe detect a change and restart all daemon docker >< | 08:44 |
| ktibi | and boom ! | 08:44 |
| ktibi | mgoddard, so https://docs.ansible.com/ansible/latest/user_guide/playbooks_delegation.html#rolling-update-batch-size :D | 08:44 |
| mgoddard | ktibi: only if you run kayobe overcloud host configure | 08:45 |
| ktibi | mgoddard, so how apply live restore ? | 08:45 |
| ktibi | host upgrade ? | 08:45 |
| mgoddard | host configure | 08:45 |
| ktibi | yes so kayobe will restart all docker daemon ? | 08:45 |
| mgoddard | but that's not in the standard upgrade procedure | 08:45 |
| mgoddard | you can apply the setting by reloading docker rather than restarting: https://docs.docker.com/config/containers/live-restore/ | 08:46 |
| mgoddard | unsure how we decide when to reload vs. restart | 08:47 |
| ktibi | https://github.com/openstack/kayobe/blob/master/ansible/roles/docker/tasks/config.yml#L7 | 08:48 |
| ktibi | hum I don't know why kayobe restart all my docker >< | 08:49 |
| ktibi | or it's kolla I guess | 08:49 |
| mgoddard | could also be kolla | 08:50 |
| mgoddard | need to check the logic in bootstrap-servers | 08:50 |
| ktibi | I change docker_resgitry here | 08:51 |
| ktibi | https://github.com/openstack/kolla-ansible/blob/stable/pike/ansible/roles/baremetal/templates/docker_systemd_service.j2#L4 | 08:51 |
| ktibi | and https://github.com/openstack/kolla-ansible/blob/stable/pike/ansible/roles/baremetal/tasks/post-install.yml#L51 | 08:52 |
| ktibi | so it's kolla | 08:52 |
| *** gkadam has quit IRC | 08:52 | |
| ktibi | I'll open a bug I think | 08:53 |
| *** gkadam has joined #openstack-kayobe | 08:53 | |
| ktibi | ok first test works, I stop daemon and all containers are alive :D | 08:57 |
| ktibi | mgoddard, so I enabled by default in kayobe and we need to make a reload or a restart but with update size feature of ansible | 08:59 |
| mgoddard | ktibi: you can't do serial per-task, but there are some workarounds that could be considered here: https://github.com/ansible/ansible/issues/12170 | 09:08 |
| mgoddard | ktibi: the daemon.json could be parsed before & after, then check for changed fields that support reload | 09:12 |
| ktibi | why can't add serial on https://github.com/openstack/kayobe/blob/master/ansible/roles/docker/tasks/config.yml#L2 ???? | 09:12 |
| ktibi | hum ho you call handler so all handler will call after the play :/ | 09:13 |
| mgoddard | you can't add serial to a task, only a play | 09:14 |
| mgoddard | we could run the play serially, will just be a bit slower | 09:14 |
| mgoddard | slow > unsafe | 09:14 |
| mgoddard | using reload would be nice though if possible | 09:15 |
| mgoddard | we only need to ensure we reload to add live-restore - after that we can restart | 09:15 |
| ktibi | I think when you change the storage of docker we need restart :/ | 09:15 |
| ktibi | is it possible to make that with ansible ? | 09:16 |
| openstackgerrit | Kevin Tibi proposed openstack/kayobe master: Support docker live-restore option https://review.openstack.org/559292 | 09:17 |
| mgoddard | yes | 09:18 |
| mgoddard | ktibi: anything's possible | 09:18 |
| mgoddard | see my comment earlier about a JSON diff | 09:19 |
| ktibi | mgoddard, where is your comment ? | 10:01 |
| mgoddard | ktibi: the daemon.json could be parsed before & after, then check for changed fields that support reload | 10:02 |
| *** dasTor has quit IRC | 10:35 | |
| *** gkadam has quit IRC | 10:47 | |
| openstackgerrit | Mark Goddard proposed openstack/kayobe master: WIP: Use the Zuul VM as the seed https://review.openstack.org/558891 | 12:14 |
| *** egonzalez_afk has quit IRC | 13:02 | |
| *** gkadam has joined #openstack-kayobe | 13:30 | |
| ktibi | mgoddard, good to read http://www.juliosblog.com/high-performance-computing-hpc-on-openstack-a-few-recommendations/ | 14:34 |
| *** gkadam_ has joined #openstack-kayobe | 14:51 | |
| *** gkadam has quit IRC | 14:53 | |
| mgoddard | thanks ktibi: looks interesting | 14:53 |
| *** gkadam_ has quit IRC | 15:10 | |
| *** openstackgerrit has quit IRC | 15:33 | |
| *** ktibi has quit IRC | 16:11 | |
| *** mgoddard has quit IRC | 16:56 | |
| *** dougsz has quit IRC | 17:24 | |
| *** mgoddard has joined #openstack-kayobe | 17:28 | |
| *** openstackgerrit has joined #openstack-kayobe | 18:34 | |
| openstackgerrit | Mark Goddard proposed openstack/kayobe master: WIP: Use the Zuul VM as the seed https://review.openstack.org/558891 | 18:34 |
| *** mgoddard has quit IRC | 18:53 | |
| *** mgoddard has joined #openstack-kayobe | 19:31 | |
| *** mgoddard has quit IRC | 19:37 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!