| opendevreview | James Gibson proposed openstack/openstack-ansible-os_nova master: [WIP] Enable TLS for live migrations https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/815224 | 07:34 |
|---|---|---|
| *** noonedeadpunk_ is now known as noonedeadpunk | 10:36 | |
| opendevreview | Andrew Bonney proposed openstack/openstack-ansible master: WIP: Add external IPv6 addressing for keepalived https://review.opendev.org/c/openstack/openstack-ansible/+/815284 | 11:25 |
| kleini | I have nova_nova_conf_overrides: actually in two different group_vars files. Unfortunately these two group_vars files apply to the same nova container. Is there a way to merge both overrides? | 11:29 |
| opendevreview | Merged openstack/openstack-ansible-os_heat stable/victoria: Fix _service_users for multi-domain deployments https://review.opendev.org/c/openstack/openstack-ansible-os_heat/+/813942 | 11:51 |
| opendevreview | Merged openstack/ansible-role-python_venv_build stable/stein: Set centos-7 jobs to non voting https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/814775 | 12:09 |
| opendevreview | Merged openstack/ansible-role-python_venv_build stable/stein: Workaround distro provided pip having old CA certs on centos-7 https://review.opendev.org/c/openstack/ansible-role-python_venv_build/+/814559 | 12:09 |
| kleini | answer to my own question: merge by hand https://paste.opendev.org/show/810191/ | 12:48 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible master: Remove reference to deprecated python-ceilometerclient https://review.opendev.org/c/openstack/openstack-ansible/+/815315 | 14:33 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible-os_ceilometer master: Remove references to deprecated python-ceilomoeterclient https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/815317 | 14:34 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible-os_heat master: Remove references to deprecated python-ceilometerclient https://review.opendev.org/c/openstack/openstack-ansible-os_heat/+/815319 | 14:35 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible-os_tempest master: Remove references to deprecated python-ceilometerclient https://review.opendev.org/c/openstack/openstack-ansible-os_tempest/+/815320 | 14:36 |
| opendevreview | Jonathan Rosser proposed openstack/openstack-ansible-os_ceilometer master: Remove references to deprecated python-ceilometerclient https://review.opendev.org/c/openstack/openstack-ansible-os_ceilometer/+/815317 | 15:22 |
| opendevreview | Merged openstack/openstack-ansible stable/stein: Fetch upper constraints file with curl rather than allow pip to download it https://review.opendev.org/c/openstack/openstack-ansible/+/814560 | 15:29 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible stable/wallaby: Bump OpenStack-Ansible Wallaby https://review.opendev.org/c/openstack/openstack-ansible/+/815333 | 16:17 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible stable/victoria: Bump OpenStack-Ansible Victoria https://review.opendev.org/c/openstack/openstack-ansible/+/815336 | 16:29 |
| mgariepy | spatel, ovn is it stable and working mostly painlessly on the latest w release? | 16:59 |
| spatel | Yes.. i believe | 16:59 |
| spatel | i had no issue so far | 17:00 |
| mgariepy | ok | 17:05 |
| mgariepy | i want to tests revovery time in case of failure with lxb, ovs and ovn. | 17:06 |
| spatel | recovery time? | 17:16 |
| mgariepy | in case of a node going down. | 17:17 |
| mgariepy | or something else happening. | 17:17 |
| spatel | oh! | 17:17 |
| mgariepy | i also want to test to see if the ha-router issue i saw in older release is fixed or not. | 17:18 |
| spatel | if cluster in dirty state i would nuke and rebuild and run neutron recovery script | 17:18 |
| spatel | i mostly use DVR gw so not a single node failure | 17:18 |
| spatel | I also configure centralized gw and check failover which is super fast because of BFD monitoring | 17:25 |
| spatel | so all and all things are good but it would be great if someone else also cross check | 17:25 |
| mgariepy | well that's what i expect a well | 17:32 |
| mgariepy | i also want to see if i can tweak a bit le lxb to make stuff faster a bit. | 17:32 |
| mgariepy | spatel, with dvr on ovn. the FIP will be on the hosting compute i guess the router will be hosted where ? | 17:33 |
| spatel | FIP will be on compute node when using DVR (all your traffic go out directly from your compute node) i did blog that also | 17:35 |
| spatel | https://satishdotpatel.github.io/openstack-ansible-ovn-deployment-part2/ | 17:35 |
| mgariepy | i do have a short memory lol. | 17:36 |
| spatel | :) we all does when lots of shit hit the fan | 17:36 |
| spatel | enable_distributed_floating_ip = True | 17:37 |
| spatel | legacy DVR consume lots of public IP because of distributed FIP but OVN DVR using openflow which do some magic and doesn't eat public IP for FIP | 17:38 |
| spatel | OVN DVR is win win for larger cloud using lots of public IPs | 17:38 |
| mgariepy | legacy dvr needed more than 1 ip for the dvr part ? | 17:41 |
| mgariepy | O_o | 17:41 |
| spatel | legacy DVR reserve 1 public IP for each compute node for DNAT | 17:41 |
| mgariepy | ho. | 17:42 |
| mgariepy | it's sad. | 17:42 |
| mgariepy | lol | 17:42 |
| spatel | if you have 200 compute nodes in your legacy cluster then you need 200 public IP just for DVR router. so total waste of public IP | 17:42 |
| spatel | legacy DVR cluster i meant | 17:42 |
| mgariepy | yep i got it. | 17:43 |
| mgariepy | do you still have network server on the control plane with ovn ? | 17:43 |
| spatel | OVN using openflow and it do some magic like anycast gateway so same public gateway IP available on all compute nodes. | 17:43 |
| spatel | network server? | 17:44 |
| mgariepy | server/service. | 17:45 |
| mgariepy | like with lxb i do have agent running on the controllers. | 17:45 |
| spatel | no | 17:45 |
| mgariepy | with ovn all that magic is done on the compute. | 17:45 |
| spatel | my ovn deployment is all DVR based | 17:46 |
| spatel | yes none of VM traffic hitting controller nodes.. all DVR | 17:46 |
| mgariepy | if there are no public FIP for the vm the router get scheduled on the compute also ? | 17:46 |
| spatel | yes | 17:47 |
| spatel | it will any compute node for gw and use for SNAT | 17:47 |
| spatel | you can dedicatedly create network node for GW.. | 17:47 |
| spatel | i like DVR so no single point of bottleneck so default i use DVR | 17:48 |
| mgariepy | yep, | 17:48 |
| mgariepy | there is not much point of doing dedicated network host if going in full DVR mode | 17:49 |
| spatel | Yes also easy for troubleshooting because you know your VM gateway is your own compute node | 17:50 |
| spatel | ansible-kolla default use OVN for deployment :) | 17:53 |
| spatel | may be one day we do same | 17:53 |
| mgariepy | yep i really thing that ovn is the future. | 17:53 |
| mgariepy | tripleO also uses ovn | 17:54 |
| mgariepy | are someone here uses maas? | 17:54 |
| mgariepy | god it's buggy .. lol | 17:55 |
| spatel | General question what is best solution for VM disk shared storage? | 19:31 |
| spatel | NFS or Ceph ? | 19:31 |
| spatel | I know ceph is best way to go when it comes to vm storage for live migration etc.. but if someone asking use NFS for less complexity. in that case any performance issue? | 19:32 |
| mgariepy | ceph does cost more but is way better. | 19:33 |
| spatel | agreed.. | 19:33 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!