| *** d34dh0r53 has quit IRC | 00:09 | |
| *** d34dh0r53 has joined #openstack-ansible | 00:12 | |
| *** cshen has quit IRC | 00:25 | |
| *** gyee has quit IRC | 00:52 | |
| *** cshen has joined #openstack-ansible | 00:52 | |
| *** also_stingrayza has joined #openstack-ansible | 01:23 | |
| *** rh-jelabarre has quit IRC | 01:23 | |
| *** stingrayza has quit IRC | 01:25 | |
| *** cshen has quit IRC | 01:36 | |
| *** spatel has joined #openstack-ansible | 01:55 | |
| openstackgerrit | ws proposed openstack/openstack-ansible master: Add rsync package for New CentOS-8 minimal installs https://review.opendev.org/740843 | 02:01 |
|---|---|---|
| *** cshen has joined #openstack-ansible | 02:04 | |
| *** spatel has quit IRC | 02:34 | |
| *** spatel has joined #openstack-ansible | 02:51 | |
| *** cshen has quit IRC | 02:54 | |
| *** schwicht has quit IRC | 03:09 | |
| *** schwicht has joined #openstack-ansible | 03:15 | |
| *** schwicht has quit IRC | 03:20 | |
| *** cshen has joined #openstack-ansible | 03:22 | |
| *** arkan has joined #openstack-ansible | 03:30 | |
| *** schwicht has joined #openstack-ansible | 03:51 | |
| *** arkan has quit IRC | 03:51 | |
| *** miouge36253686 has quit IRC | 03:54 | |
| *** mbuil has quit IRC | 03:54 | |
| *** miouge362536868 has joined #openstack-ansible | 03:54 | |
| *** mbuil has joined #openstack-ansible | 03:55 | |
| *** schwicht has quit IRC | 04:06 | |
| *** cshen has quit IRC | 04:19 | |
| *** evrardjp has quit IRC | 04:33 | |
| *** evrardjp has joined #openstack-ansible | 04:33 | |
| *** cshen has joined #openstack-ansible | 04:51 | |
| *** arkan has joined #openstack-ansible | 05:03 | |
| *** spatel has quit IRC | 05:09 | |
| janno | jrosser: no worries, thank you | 05:28 |
| *** udesale has joined #openstack-ansible | 05:29 | |
| *** cshen has quit IRC | 05:46 | |
| *** arkan has quit IRC | 05:51 | |
| openstackgerrit | Chandan Kumar (raukadah) proposed openstack/openstack-ansible master: [WIP] improve ironic tempest testing https://review.opendev.org/736507 | 05:59 |
| openstackgerrit | Chandan Kumar (raukadah) proposed openstack/openstack-ansible master: [WIP] improve ironic tempest testing https://review.opendev.org/736507 | 05:59 |
| *** arkan has joined #openstack-ansible | 06:16 | |
| *** cshen has joined #openstack-ansible | 06:19 | |
| *** halali_ has quit IRC | 06:30 | |
| *** halali_ has joined #openstack-ansible | 06:51 | |
| *** this10nly has joined #openstack-ansible | 06:55 | |
| *** cshen has quit IRC | 06:57 | |
| *** cshen has joined #openstack-ansible | 07:13 | |
| *** halali_ has quit IRC | 07:34 | |
| *** tosky has joined #openstack-ansible | 07:37 | |
| *** arkan has quit IRC | 07:51 | |
| *** arkan has joined #openstack-ansible | 07:52 | |
| *** donnyd has quit IRC | 08:53 | |
| *** donnyd has joined #openstack-ansible | 08:53 | |
| *** also_stingrayza is now known as stingrayza | 08:58 | |
| *** cshen has quit IRC | 08:59 | |
| ioni | hi guys, what's the procedure to update ceph ? i've updated an aio system from train nto ussuri and ceph wasn't updated on mons | 09:12 |
| ioni | it was updated only on osds | 09:12 |
| jrosser | ioni: this may be useful https://review.opendev.org/#/c/711440/ | 09:21 |
| ioni | jrosser: ok, i'll apply the patch and retry | 09:23 |
| ioni | jrosser: is useful for all branches not only for stable/stein | 09:24 |
| *** cshen has joined #openstack-ansible | 09:26 | |
| jrosser | ioni: if you can check that works we should look to merge and backport it | 09:29 |
| *** soren has quit IRC | 10:01 | |
| *** soren has joined #openstack-ansible | 10:03 | |
| *** grantza has quit IRC | 10:04 | |
| *** cshen has quit IRC | 10:08 | |
| ioni | jrosser: seems to work, i only used the upgrade yml | 10:09 |
| *** bl0m1 has quit IRC | 10:22 | |
| *** gouthamr has quit IRC | 10:22 | |
| *** gouthamr has joined #openstack-ansible | 10:23 | |
| *** bl0m1 has joined #openstack-ansible | 10:23 | |
| *** Jeffrey4l has quit IRC | 10:23 | |
| *** Jeffrey4l has joined #openstack-ansible | 10:24 | |
| *** cshen has joined #openstack-ansible | 10:38 | |
| *** cshen has quit IRC | 10:43 | |
| openstackgerrit | Merged openstack/ansible-role-systemd_service stable/ussuri: Remove fail: tasks https://review.opendev.org/739566 | 10:51 |
| *** arkan has quit IRC | 10:51 | |
| *** gokhani has joined #openstack-ansible | 10:51 | |
| *** cshen has joined #openstack-ansible | 11:18 | |
| *** arkan has joined #openstack-ansible | 11:19 | |
| *** cshen has quit IRC | 11:22 | |
| *** arkan has quit IRC | 11:52 | |
| *** arkan has joined #openstack-ansible | 11:52 | |
| mgariepy | morning everyone | 11:55 |
| *** spatel has joined #openstack-ansible | 11:58 | |
| *** schwicht has joined #openstack-ansible | 12:03 | |
| *** spatel has quit IRC | 12:03 | |
| *** rh-jelabarre has joined #openstack-ansible | 12:09 | |
| *** udesale_ has joined #openstack-ansible | 12:22 | |
| *** udesale has quit IRC | 12:25 | |
| *** cshen has joined #openstack-ansible | 12:53 | |
| *** spatel has joined #openstack-ansible | 12:58 | |
| *** cshen has quit IRC | 13:03 | |
| jrosser | mgariepy: CeeMac are either of you using a network thats not br-mgmt as the hosts in /etc/openstack_deploy/openstack_user_config.yml ? | 13:30 |
| jrosser | like a dedicated provisioning network | 13:30 |
| *** pcaruana has quit IRC | 13:31 | |
| *** pcaruana has joined #openstack-ansible | 13:43 | |
| spatel | morning folks | 13:48 |
| spatel | jrosser: you may help with this - https://review.opendev.org/#/c/740513/ | 13:49 |
| spatel | This patch is totally good but somehow build worker not picking up correctly and still showing same error | 13:50 |
| spatel | it seem build workers are having issue.. | 13:50 |
| CeeMac | Jrosser yes I use an OOB management network for my hosts connections | 13:51 |
| *** arkan has quit IRC | 13:51 | |
| jrosser | CeeMac: i wonder if you can do a sanity check for me | 13:53 |
| jrosser | we are just looking at this here https://opendev.org/openstack/openstack-ansible/src/branch/master/playbooks/common-playbooks/nova.yml#L69-L74 | 13:53 |
| jrosser | that does some magical stuff rooting round in the hostvars to pick an address for the nova host | 13:53 |
| jrosser | it then sticks it here https://opendev.org/openstack/openstack-ansible/src/branch/master/playbooks/common-playbooks/nova.yml#L145 | 13:54 |
| jrosser | which means that IP ends up in places like this https://opendev.org/openstack/openstack-ansible-os_nova/src/branch/master/templates/nova.conf.j2#L41 | 13:55 |
| jrosser | would be interesting to see if you have the IP you expect to have there in nova.conf | 13:55 |
| CeeMac | So that's management as in br-mgmt as opposed to the host local management address? | 13:55 |
| jrosser | that is a very good question :) | 13:56 |
| CeeMac | I'm not at my computer right now but I'll double check when I get back to my desk later. Working split shifts atm | 13:56 |
| jrosser | ah ok, no worries | 13:56 |
| jrosser | we are thinking that "management_address" is wrong and does not exist in container_networks in hostvars | 13:57 |
| CeeMac | I'll be back online by 1700 latest | 13:57 |
| jrosser | there is a fallback which would be ansible_host | 13:58 |
| jrosser | so the address of the OOB network ends up in places that it maybe shouldnt | 13:58 |
| *** dave-mccowan has joined #openstack-ansible | 14:02 | |
| *** cshen has joined #openstack-ansible | 14:08 | |
| CeeMac | Has anything changed in how this works since rocky or should it still be the same | 14:17 |
| *** this10nly has quit IRC | 14:26 | |
| *** cshen has quit IRC | 14:27 | |
| mgariepy | jrosser, not me. | 14:31 |
| *** dave-mccowan has quit IRC | 14:33 | |
| jrosser | CeeMac: it should be the same | 14:35 |
| CeeMac | Ok | 14:39 |
| *** dave-mccowan has joined #openstack-ansible | 14:39 | |
| *** cshen has joined #openstack-ansible | 14:54 | |
| *** arkan has joined #openstack-ansible | 15:00 | |
| jrosser | spatel: the issue with your patch to gnocchi is that the test fails when installing ceilometer, becasue this is not merged https://review.opendev.org/#/c/739400/ | 15:09 |
| jrosser | just here the error is in os_ceilometer https://zuul.opendev.org/t/openstack/build/6424de617c0240719b18a44eb350bae1/log/job-output.txt#13643 | 15:10 |
| spatel | jrosser: you are saying its dependency and without merge it won't work? | 15:10 |
| jrosser | it is a dependacy yes, for the gnocci role see the test scenario here https://zuul.opendev.org/t/openstack/build/6424de617c0240719b18a44eb350bae1/log/job-output.txt#5238 | 15:12 |
| jrosser | 'telemetry' means it deploys the whole telemetry stack | 15:12 |
| spatel | hmm that make sense | 15:13 |
| jrosser | i do not think we should be too concerned about the peripheral things like this | 15:13 |
| jrosser | we should get centos-8 basic stuff passing on ussuri first | 15:14 |
| spatel | jrosser: yes, also this patch is very minor change just rename of rehat-7.yml file and in some case just package name match for centos8 | 15:14 |
| spatel | i think its totally safe to merge if people review them | 15:14 |
| jrosser | you can't :) | 15:14 |
| jrosser | unless the tests pass it will never merge | 15:15 |
| jrosser | or the tests are set to not-voting | 15:15 |
| jrosser | there is no "merge now regardless" | 15:15 |
| spatel | in this case what do you suggest? | 15:15 |
| spatel | also we are very close to centos-8 beta release, as i am seeing all major patches has been merged | 15:17 |
| jrosser | so two different things here | 15:17 |
| jrosser | your test is failing for gnocchi becasue os_ceilometer patch is not yet merged | 15:18 |
| jrosser | and to merge the os_ceilometer patch we need to merge the gnocchi one | 15:18 |
| spatel | :) chicken egg | 15:18 |
| jrosser | so that is clearly an impossible situation and the only was to resolve that is to set the centos-8 tests to not voting and merge some of the patches with a failing test | 15:18 |
| jrosser | *but* | 15:18 |
| jrosser | you can still see beforehand if it is going to work | 15:19 |
| jrosser | for example, you can make the os_ceilometer patch also take account of the unmerged patch to gnocchi | 15:19 |
| jrosser | make it do a "what if" test | 15:20 |
| jrosser | here is an unrelated example https://review.opendev.org/#/c/618305/ | 15:20 |
| jrosser | see the 'Depends-On' line in the commit message, this means two things | 15:20 |
| jrosser | "do not allow this patch to merge before the depends-on one is merged" | 15:21 |
| jrosser | "when testing this patch also apply the depends-on patches in the test" | 15:21 |
| jrosser | so if you edit the commit message of the os_ceilometer patch to include depends-on lines for the other parts of the telemetry stack we will see further on how far it works | 15:22 |
| spatel | hmm! interesting.. let me see | 15:23 |
| spatel | jrosser: thanks, let me try and i will ping you back if any question | 15:24 |
| jrosser | sure no worries | 15:24 |
| jrosser | this is one of the powerful bits of testing with zuul that you can make a change in one repo and test the effect that would have in another, before merging anything | 15:25 |
| arkan | Hi guys, how are you all ? I hope you are doing well. | 15:26 |
| arkan | I got a hint from @CeeMac to use magnum in stead of using terraform + kubespray, so I went and installed it :)) | 15:26 |
| arkan | with barbican also | 15:26 |
| arkan | does coreos works with magnum (stable/train) ? | 15:28 |
| arkan | I've created the template successfully on horizon | 15:28 |
| arkan | and also on cli | 15:28 |
| arkan | only the cluster when it is in creation in progress | 15:29 |
| arkan | then it will through this error | 15:29 |
| arkan | ERROR: The Parameter (fixed_network_name) was not defined in template. | 15:29 |
| arkan | and I have specified it in my template | 15:29 |
| arkan | from horizon | 15:30 |
| arkan | there are combo boxes for both "Fixed Network" and "Fixed Subnet" in the Network section | 15:31 |
| *** udesale_ has quit IRC | 15:35 | |
| arkan | humm | 15:38 |
| arkan | maybe this is have to do with coreos | 15:38 |
| arkan | i will try fedora-atomic | 15:38 |
| arkan | I will need to download it first | 15:39 |
| *** gyee has joined #openstack-ansible | 15:50 | |
| *** cshen has quit IRC | 15:50 | |
| admin0 | is adding magnum_infra_hosts: and then rerunning bootstrap/setup is enough to get magnum support for deploying kubernetes ? | 15:52 |
| *** spatel has quit IRC | 15:54 | |
| CeeMac | jrosser: in nova.conf my_ip is set to the OOB management IP | 15:54 |
| jrosser | CeeMac: i'm not completely clear on the detail of why just yet but the upshot is that live migration seems to happen across that network | 15:56 |
| jrosser | we had some migrations that took eternity to complete and that was why | 15:57 |
| CeeMac | oh | 16:04 |
| CeeMac | well, thats not goodf | 16:04 |
| jrosser | we've not fully understood whats happening as the nova docs are a bit confusing | 16:06 |
| *** sshnaidm is now known as sshnaidm|afk | 16:08 | |
| CeeMac | it should be using br-mgmt presumably? | 16:08 |
| jrosser | yes so it starts here https://github.com/openstack/openstack-ansible/blob/master/playbooks/common-playbooks/nova.yml#L69-L74 | 16:12 |
| jrosser | and that takes the dict container_networks for the host, and tries to find the key "managment_address" in there | 16:13 |
| jrosser | but i think that is wrong as i've not found good reference to managment_address in the code or documentation | 16:14 |
| jrosser | that would come about by having a section "management" in cidr_networks in openstack_user_config.yml which none of the examples have | 16:14 |
| CeeMac | yes, i was just looking at dynamic-address-fact.yml trying to match it back | 16:17 |
| CeeMac | is not pulling the "management_address" value out of the query in dynamic-address-fact.yml | 16:19 |
| CeeMac | which is looking for "\"ansible_{{ hostvars[inventory_hostname] in metal_query? | 16:20 |
| jrosser | if you do ansible -m debug -a 'var=container_networks' then i don't think you'll find a section in the dict for managment_address | 16:21 |
| jrosser | would be interesting to see if you do | 16:21 |
| jrosser | ansible <host> -m debug .... | 16:22 |
| CeeMac | ok | 16:22 |
| CeeMac | was just looking in openstack_inventory and the IP is references in ansible_host and container_address and container_networks: container_address: address | 16:23 |
| *** cshen has joined #openstack-ansible | 16:23 | |
| *** aedc_ has joined #openstack-ansible | 16:24 | |
| *** aedc_ has quit IRC | 16:24 | |
| CeeMac | no its definitely container_address now management_address | 16:25 |
| arkan | admin0: setup-hosts, setup-infrastructure, os-horizon, os-magnum | 16:25 |
| arkan | admin0: it has problem with coreos | 16:25 |
| jrosser | CeeMac: for a metal host the query should look at the "bridge" field | 16:26 |
| jrosser | then take the ip from the bridge of that name | 16:26 |
| CeeMac | ah, i couldn't work it if it was using | as 'or' in an order of priority | 16:26 |
| CeeMac | but i dont have my management interface in a bridge.... | 16:27 |
| *** aedc has quit IRC | 16:27 | |
| arkan | admin0: now the problem disappeared when I used fedora-atomic-27 image but another problem I have now in magnum container: Failed to load default keystone auth policy: FileNotFoundError: [Errno 2] No such file or directory: '/etc/magnum/keystone_auth_default_policy.json' | 16:27 |
| CeeMac | json_query(find_bridge) <- that bit? | 16:27 |
| jrosser | :) br-mgmt is kind of expected i think | 16:27 |
| CeeMac | right | 16:28 |
| jrosser | but also then just look here https://docs.openstack.org/nova/latest/configuration/config.html | 16:28 |
| CeeMac | except it isnt, hmm | 16:28 |
| *** cshen has quit IRC | 16:28 | |
| jrosser | at the definition of live_migration_uri | 16:28 |
| jrosser | then we all collectively go "oh no!" | 16:28 |
| jrosser | CeeMac: anyway just good to see that for you with a similar setup it also makes the same outcome as us | 16:29 |
| jrosser | we're going to have to (tomorrow) dive into the nova code and see really how that migration uri is used | 16:29 |
| CeeMac | it definitely doesn't read well | 16:30 |
| CeeMac | If live_migration_inbound_addr value is not None and live_migration_tunnelled is False, the ip/hostname address of target compute node is used instead of live_migration_uri as the uri for live migration. | 16:30 |
| CeeMac | and the hosts file is built from the inventory/management address, in my case the OOB one | 16:31 |
| CeeMac | jrosser: just looking at openstack_inventory again | 16:34 |
| CeeMac | container_address has the OOB management IP but br-mgmt as the bridge | 16:35 |
| jrosser | yes, same here | 16:35 |
| CeeMac | so looking for br-mgmt would return the wrong IP in that case | 16:35 |
| jrosser | so for us we would get part of the way there by changing "management_address" (confusingly) to "container_address" | 16:35 |
| jrosser | in this case for nova i think the ip off br-mgmt is the thing you want it to be using | 16:36 |
| CeeMac | but then why isn't the dynamic inventory loading the br-mgmt IP as the container-address | 16:36 |
| jrosser | for the compute host? | 16:36 |
| CeeMac | yes | 16:36 |
| jrosser | i see something like this http://paste.openstack.org/show/795918/ | 16:38 |
| CeeMac | the the ansible_host is the OOB mgmt IP if you have similar setup to me | 16:39 |
| jrosser | only for the things that are hosts though, the 10.84 ones | 16:40 |
| CeeMac | ? | 16:40 |
| CeeMac | got you | 16:40 |
| jrosser | 10.80 is the openstack mgmt network | 16:40 |
| jrosser | which kind of makes sense for ansible_host | 16:40 |
| CeeMac | in the 'standard' way of deploying yes | 16:40 |
| jrosser | but not so much when the dynamic address lookup tries to lookup something non-existing (in any deployment) and falls back to ansible_host | 16:41 |
| jrosser | normally this would be OK as mgmt address == ansible address | 16:41 |
| CeeMac | however, with management_bridge: "br-mgmt" set would it not gather the br-mgmt ip for the container address? | 16:41 |
| jrosser | i'm not sure that variable is actually doing anything though | 16:42 |
| CeeMac | i see, so potentially its falling back to "non_metal_query: "container_networks.{{ network_address }}.address" first | 16:42 |
| CeeMac | which in our case is populated by the OOB IP not the br-mgmt IP for metal hosts | 16:43 |
| *** cshen has joined #openstack-ansible | 16:43 | |
| CeeMac | actually, i missed the find_bridge bit | 16:44 |
| jrosser | yeah :) it's mind boggling isnt it | 16:44 |
| CeeMac | saying that, I don't think the issue is necessarily in dynamic-address-fact.yml? | 16:45 |
| CeeMac | if the wrong IP is in the dynamic inventory for the mr-mgmt bridge in the first place? | 16:45 |
| CeeMac | *br-mgmt | 16:45 |
| jrosser | the issue is that its trying to find a somthing called "management_address" | 16:45 |
| jrosser | which will never exist | 16:46 |
| CeeMac | i still haven't quite worked that bit out :) | 16:46 |
| CeeMac | ok | 16:46 |
| CeeMac | now i have | 16:46 |
| jrosser | here https://github.com/openstack/openstack-ansible/blob/master/playbooks/common-playbooks/nova.yml#L72 | 16:46 |
| CeeMac | so its replacing {{ network_address }} with management_address in the queries | 16:46 |
| CeeMac | i was trying to work out which way it was passing the variable | 16:47 |
| CeeMac | im a bit rusty :) | 16:47 |
| jrosser | yes its setting a fact "management_address" by looking up "management_address" in container_networks then performing the jq on it to get the address | 16:47 |
| jrosser | which is umm, 'esoteric' | 16:48 |
| CeeMac | which, as you say, isn't going to get anywhere | 16:48 |
| CeeMac | good word :) | 16:48 |
| *** sshnaidm|afk is now known as sshnaidm | 16:48 | |
| jrosser | right - so i think this works for standard deploys becasue it falls back to the right thing | 16:48 |
| jrosser | sort of by accident | 16:48 |
| CeeMac | yes | 16:49 |
| CeeMac | but using the ansible_host value | 16:49 |
| CeeMac | s/but/by | 16:49 |
| jrosser | right | 16:49 |
| CeeMac | as the rest of the queries are going to fail | 16:49 |
| jrosser | we'll dig into this a bit more, becasue it seems that getting the value you want in my_ip isnt enough to steer the migration traffic to br-mgmt | 16:50 |
| jrosser | *tomorrow | 16:50 |
| *** arkan has quit IRC | 16:51 | |
| CeeMac | yes | 16:51 |
| CeeMac | thats something else again then | 16:51 |
| CeeMac | but those live_migration vars are pretty out there too | 16:51 |
| *** arkan has joined #openstack-ansible | 16:52 | |
| openstackgerrit | ws proposed openstack/openstack-ansible master: Add rsync package for New CentOS-8 minimal installs https://review.opendev.org/740843 | 16:56 |
| CeeMac | ok, so my_ip is set by nova_management_address which is set from management_address | 16:58 |
| CeeMac | interesting | 16:59 |
| CeeMac | jrosser: https://opendev.org/openstack/openstack-ansible-os_nova/src/branch/master/tests/group_vars/all_containers.yml#L17 | 16:59 |
| *** npalladium has joined #openstack-ansible | 17:03 | |
| CeeMac | so nova.conf.j2 sets live_migration_tunnelled = true, and configures live_migration_uri. | 17:15 |
| CeeMac | which means that live_migration_inbound_addr won't be used | 17:15 |
| CeeMac | but %s replaced with the migration target hostname, which in our case will resolve to the OOB management address | 17:15 |
| CeeMac | but then what could you use to replace %s with that would resolve the br-mgmt ip dynamically as part the uri buildout | 17:17 |
| CeeMac | oh, actually hosts has an entry for both the OOB IP and br-mgmt | 17:19 |
| CeeMac | so thats going to be confusing | 17:19 |
| CeeMac | is that going to round-robin on resolution requests? | 17:20 |
| jrosser | i am wondering if using %s is a bad idea | 17:24 |
| jrosser | becasue it makes some assumptions about what dns may or may not say | 17:24 |
| CeeMac | yes | 17:24 |
| jrosser | it might be better to template out the actual IP that we want then there is no confusion | 17:25 |
| CeeMac | but what could you replace it with that would be dynamic across target migration hosts? | 17:25 |
| jrosser | {{ nova_managment_address }} i guess, if that could be fixed to be the right thing | 17:25 |
| CeeMac | as i assume the URL would change based on where you are migrating to? | 17:25 |
| jrosser | i think its the "from" | 17:25 |
| CeeMac | this isn't proxied via haproxy? | 17:26 |
| jrosser | to the "to" would connect to this, but this is kind of where my understanding is running out | 17:26 |
| jrosser | its an ssh | 17:26 |
| CeeMac | "qemu+ssh://nova@%s/system?no_verify=1&keyfile={{ nova_system_home_folder }}/.ssh/id_rsa" | 17:26 |
| CeeMac | is what it is being set to | 17:26 |
| jrosser | https://github.com/openstack/openstack-ansible-os_nova/blob/master/templates/nova.conf.j2#L236 | 17:26 |
| CeeMac | which is a destination path as I understand it | 17:26 |
| CeeMac | yes | 17:27 |
| jrosser | oh right | 17:27 |
| CeeMac | "Override the default libvirt live migration target URI" | 17:27 |
| CeeMac | hence the conundrum | 17:27 |
| CeeMac | as that needs to be explicitly set for the destination host in the migration request | 17:28 |
| jrosser | so i figure that what matters is from the "initiator" or whatever we want to call, reaching that URI uses the right interface | 17:28 |
| CeeMac | yes, based on what address the destination host address resolves to | 17:28 |
| jrosser | hrrm, i'm a bit confused about it really | 17:29 |
| jrosser | if that uri and %s is used every time a migration is done, or else it's saying "i am this" | 17:29 |
| CeeMac | you could hack it by adding a unique hosts entry based on <hostname>_uri and setting that to the correct ip, but will hacking that into the uri work with %s_url | 17:30 |
| CeeMac | "Any included “%s” is replaced with the migration target hostname." | 17:30 |
| CeeMac | i think its set at the migration creation | 17:30 |
| *** cshen has quit IRC | 17:31 | |
| jrosser | live_migration_inbound_addr sounds like a much more useful construct | 17:32 |
| jrosser | becasue it's data about the host itself | 17:32 |
| CeeMac | you'd have to unset live_migration_tunnelled then | 17:33 |
| CeeMac | or do you mean wrt my hack? | 17:34 |
| CeeMac | ok, i'm catching up on your confusion now | 17:35 |
| CeeMac | one sec | 17:36 |
| *** mgariepy has quit IRC | 17:36 | |
| CeeMac | ok so its actually written in to nova.conf with the %s too | 17:37 |
| CeeMac | so, i guess the question is, is live_migration_uri on compute1 used to work out the correct path to a migration target | 17:37 |
| CeeMac | or is it provided to the compute node initiating the migration in a "send traffic here" kind of way | 17:38 |
| jrosser | oh my goodness https://review.opendev.org/#/c/456571/ | 17:39 |
| CeeMac | live_migration_uri is actually deprecated anyway..... | 17:39 |
| CeeMac | ah, you found something similar a different way :) | 17:40 |
| CeeMac | digging in think having tunnel enabled is still going to cause some issues | 17:46 |
| CeeMac | "It is typically only necessary to specify this URI if the destination host has multiple interfaces and a specific interface is required to transmit migration data" | 17:48 |
| CeeMac | that seems promising | 17:48 |
| CeeMac | https://libvirt.org/html/libvirt-libvirt-domain.html#VIR_MIGRATE_PARAM_URI | 17:48 |
| jrosser | i think i'm going to leave this until tomorrow | 17:48 |
| jrosser | but if you have any useful insights for when we pick it up in the morning do let me know | 17:48 |
| CeeMac | i'll be clocking off shortly too, but i'll take a peak again first thing | 17:49 |
| jrosser | ok, no worries, catch up tomorrow | 17:49 |
| CeeMac | sure, have a good evening :) | 17:49 |
| CeeMac | I'll go back to lurking | 17:50 |
| *** schwicht has quit IRC | 17:56 | |
| *** schwicht has joined #openstack-ansible | 18:02 | |
| *** cshen has joined #openstack-ansible | 18:04 | |
| *** d34dh0r53 has quit IRC | 18:08 | |
| *** cshen has quit IRC | 18:08 | |
| *** d34dh0r53 has joined #openstack-ansible | 18:15 | |
| *** mgariepy has joined #openstack-ansible | 18:22 | |
| *** rmcall has joined #openstack-ansible | 18:30 | |
| *** spatel has joined #openstack-ansible | 18:37 | |
| CeeMac | jrosser: for the morning (because apparently I'm incapable of letting things go once I'm curious). I think setting live_migration_inbound_addr to nova_management_address and setting live_migration_tunnelled to false could be the way forward. That leaves changing 'management_address' to 'container_adress', and ensuring the correct br-mgmt ip is collected on the container_networks dict for all use cases (being | 19:01 |
| CeeMac | standard collapsed mgmt and OOB mgmt mainly) | 19:01 |
| *** mgariepy has quit IRC | 19:03 | |
| *** npalladium has quit IRC | 19:12 | |
| *** cshen has joined #openstack-ansible | 20:01 | |
| *** also_stingrayza has joined #openstack-ansible | 20:03 | |
| *** cshen has quit IRC | 20:05 | |
| *** stingrayza has quit IRC | 20:07 | |
| *** arkan has quit IRC | 20:51 | |
| *** markvoelker has joined #openstack-ansible | 21:23 | |
| CeeMac | Also, need to re-read and try and understand what is happening in this block https://opendev.org/openstack/openstack-ansible/src/branch/master/osa_toolkit/generate.py#L665 | 21:26 |
| *** markvoelker has quit IRC | 21:27 | |
| *** spatel has quit IRC | 21:44 | |
| *** spatel has joined #openstack-ansible | 21:45 | |
| *** cshen has joined #openstack-ansible | 22:02 | |
| tow | hi guys, we are kind of struggling with 20.1.3, the setup-openstack deployment gets stuck at different points for the same reason, git not found inside some services containers (placement, heat-api, aodh). If we get into the containers and install it manually, then it seems to go through. Haven't found anything in the bugtracker. | 22:04 |
| *** spatel has quit IRC | 22:04 | |
| *** cshen has quit IRC | 22:07 | |
| *** spatel has joined #openstack-ansible | 22:23 | |
| *** spatel has quit IRC | 22:23 | |
| *** tosky has quit IRC | 22:50 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!