| *** arxcruz|off is now known as arxcruz | 07:50 | |
| *** dviroel|ruck|afk is now known as dviroel|ruck | 11:26 | |
| *** anbanerj is now known as frenzyfriday | 12:58 | |
| noonedeadpunk | jamesdenton: MrClayPole_ that was the fix we had for the issue you found for V https://opendev.org/openstack/openstack-ansible-os_nova/commit/f15417aa316b6b94d5fc1988df8b71ea1bb5f246 | 13:12 |
|---|---|---|
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_nova stable/victoria: Do not use imports inside included task https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/834138 | 13:12 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-os_nova stable/victoria: Do not use imports inside included task https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/834138 | 13:16 |
| noonedeadpunk | this is backport ^ | 13:16 |
| MrClayPole_ | Thanks noonedeadpunk, I'll give it a test and report back. | 13:16 |
| *** arxcruz is now known as arxcruz|ruck | 13:25 | |
| jamesdenton | appreciate that! | 13:26 |
| *** arxcruz|ruck is now known as arxcruz | 13:27 | |
| spatel | is github is down ? | 14:21 |
| spatel | hope its not running on windows servers :( | 14:22 |
| noonedeadpunk | works for me | 14:22 |
| spatel | getting -> master -> master (Internal Server Error) with pushing code | 14:23 |
| spatel | may be issue in US | 14:23 |
| spatel | https://github.com/ also not working | 14:23 |
| *** tosky_ is now known as tosky | 14:27 | |
| mac189 | i have a question about installing xena if anyone has a chance...i'm failing at the 'os-keystone-install.yml' part of 'setup-openstack.yml' | 14:35 |
| mac189 | I'm running into a problem at the part "Build wheels for the packages to be installed into the venv". it wants to use "git+{{ keystone_git_repo}}..." in the 'keystone_pip_packages variable' found in /etc/ansible/roles/os_keystone/defaults/main.yml, but it cannot clone http://<gitserver>/keystone. I suspect this is due to the fact that my gitserver is actually a gitolite server and clones are done as "git clone | 14:36 |
| mac189 | ssh://gitolite3@<gitolite_server>" I'm not sure how to manipulate the keystone_git_repo variable or the keystone_pip_packages variable so that this will work. Any suggestions? | 14:36 |
| noonedeadpunk | mac189: and how have you defined keystone_git_repo? As ssh://gitolite3@<gitolite_server>/keystone? | 14:39 |
| mac189 | or is there a way to have it work with gitlab (i.e. git@gitlab.server.com)? | 14:39 |
| noonedeadpunk | I don't think it matters? | 14:39 |
| noonedeadpunk | I mean - you can set any url for git repo, it can use ssh or http protocol | 14:40 |
| spatel | curious why do you want to use different URL? | 14:40 |
| noonedeadpunk | as long as auth can be passed, actually | 14:40 |
| * noonedeadpunk thinks it's about forked roles | 14:40 | |
| spatel | oh own repo | 14:41 |
| * noonedeadpunk thinks it's one of those who don't contribute back | 14:41 | |
| mac189 | hmm, ok...maybe i have a typo somewhere else in my config. I'm trying to generate of alist of offline repos by set_fact ofline_openstack_services_repos and replacing the http://opendev url with the one i want to use | 14:41 |
| mac189 | well we are in a disconnected environment, so have to copy all the outside repos in and then use those copies | 14:42 |
| spatel | noonedeadpunk https://www.githubstatus.com/ | 14:42 |
| noonedeadpunk | mac189: but you obviously haven't replaced https:// | 14:42 |
| mac189 | yes, as i'm typing this i realize that may be where my error is...sorry for the mistake | 14:42 |
| noonedeadpunk | mac189: and basically they're all here https://opendev.org/openstack/openstack-ansible/src/branch/master/playbooks/defaults/repo_packages/openstack_services.yml ? | 14:42 |
| noonedeadpunk | well, not taking into account roles and collections | 14:43 |
| noonedeadpunk | mac189: um, and you can't have proxy to reach at least opendev.org? | 14:43 |
| mac189 | no :( | 14:43 |
| noonedeadpunk | well... and you already have local mirror of pypi? | 14:44 |
| noonedeadpunk | actually for `ansible-role-requirements.yml` you can have /etc/openstack_deploy/user_role_requirements.yml that should have prescedence over default one | 14:45 |
| mac189 | ah ok, thanks | 14:45 |
| noonedeadpunk | same applies for collection requirements | 14:50 |
| jrosser | pypi mirror is no fun at all | 14:51 |
| noonedeadpunk | I kind of wonder if we still have option to configure pip.conf? | 14:51 |
| noonedeadpunk | I can recall dropping pip.conf in Q | 14:51 |
| jrosser | i would say that if you are in a disconnected environment then your host provisioning should already be responsible to doing $stuff to /etc/pip.conf | 14:54 |
| jrosser | it's not really an OSA concern at that point? | 14:54 |
| jrosser | mac189: did you make a pip mirror? | 14:56 |
| mac189 | the pip mirror we have is something i don't have access to in order to maintain...but it looks like it isn't as up to date as it needs to be so i'm going to reach out to those responsible for it | 14:57 |
| jrosser | ah ok, thats good | 14:58 |
| mac189 | thanks, i was so hung up in the "git clone" part of the error I was seeing that I missed the parts about pip | 14:58 |
| mac189 | i appreciate the help and guidance with that | 14:59 |
| jrosser | you *should* be OK - you might find other things like hatop that it tries to download from the internet and you need to provide a local version | 15:00 |
| jrosser | and set things like haproxy_hatop_download_url | 15:00 |
| noonedeadpunk | jrosser: but do we copy pip.conf to container? | 15:01 |
| jrosser | based on my deployment i'd say yes | 15:02 |
| jrosser | but would have to check where/how thats happened | 15:02 |
| noonedeadpunk | ah, you need to set lxc_container_cache_files_from_host explicitly for that | 15:03 |
| noonedeadpunk | https://opendev.org/openstack/openstack-ansible/src/branch/master/tests/roles/bootstrap-host/templates/user_variables.aio.yml.j2#L247 | 15:03 |
| jrosser | i have this https://paste.opendev.org/show/bEFxwgvFSfzahap8pdz2/ | 15:04 |
| mac189 | ah shoot, i do not have that variable override...thank you! | 15:05 |
| jrosser | https://paste.opendev.org/show/bXsDdBsh8CIg09SPOwdj/ | 15:06 |
| jrosser | mac189: i think it's pretty certain you will take a few attempts at this | 15:07 |
| jrosser | completely offline install is possibly but theres a lot to deal with | 15:08 |
| mac189 | that is the truth! | 15:08 |
| MrClayPole_ | noonedeadpunk: I can confirm that the patch https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/834138 resolves my issue with nova_virt_type. | 15:18 |
| noonedeadpunk | nice! | 15:19 |
| opendevreview | Dmitriy Rabotyagov proposed openstack/openstack-ansible-lxc_hosts master: Replace systemd-mount template with role https://review.opendev.org/c/openstack/openstack-ansible-lxc_hosts/+/834183 | 15:24 |
| noonedeadpunk | if somebody was interested of the result of mysql issue I mentioned days back - it was keepalived ¯\(◉◡◔)/¯ | 15:33 |
| andrewbonney | Was that the single-region impacting issue? What about keepalived was causing it? | 15:38 |
| noonedeadpunk | So, we appeared to have multicast being routed from another region. And we had same router IDs, but different pass. It appears that pass is not respected by slaves, so slaves were recieveing announce of VRRP from other region and promoting them as master for a 2 seconds several times an hour. So we got 2 VIPs and connections that arrived on slave were dead as in second IP was removed from it. | 15:41 |
| noonedeadpunk | And master was not affected at all | 15:41 |
| noonedeadpunk | And mysql is kind of most affected thing of that because of having too much connections that goes up and down | 15:42 |
| andrewbonney | Well discovered. That sounds quite involved to debug | 15:44 |
| noonedeadpunk | well, that master keepalived showed in logs nothing gave us false direction for several days.... | 15:47 |
| noonedeadpunk | but yeah, that was interesting :) | 15:48 |
| *** dviroel|ruck is now known as dviroel|ruck|lunch | 15:56 | |
| jrosser | ah interesting, we've got haproxy_keepalived_*_virtual_router_id set to some value thats unique per deployment | 15:57 |
| jrosser | as i think we had something similar a long time ago with an AIO messing up a test lab deployment in the same subnet | 15:58 |
| noonedeadpunk | yeah and we left jsut defaults as thought that regions are fully separated... | 16:10 |
| noonedeadpunk | and it wasn't an issue until upgrade to Xena | 16:11 |
| noonedeadpunk | so upgrade to Xena likely triggered update of keepalived that raised the issue | 16:11 |
| anskiy | I'm using unicast, which looks kinda ugly in user_variables, because I have to override both whole instances :) Still haven't got time to debug my timeouts properly -_- | 16:12 |
| noonedeadpunk | anskiy: you can suggest patch to make things configurable ;) | 16:15 |
| anskiy | I was looking the other day to submit a fix for a typo in ksm tasks (which completely breaks playbook run) -- got lost in "how to contribute" :) | 16:17 |
| noonedeadpunk | anskiy: 1. register on review.openstack.org, set username and upload ssh key (and agree on some terms there). 2 install git-review 3. git clone; git commit; git review | 16:20 |
| anskiy | noonedeadpunk: that looks easy, thanks! | 16:28 |
| noonedeadpunk | so this doc for first 2 steps https://docs.openstack.org/contributors/common/setup-gerrit.html#individual-contributor-license-agreement | 16:32 |
| noonedeadpunk | that is for 3 https://docs.opendev.org/opendev/infra-manual/latest/gettingstarted.html#proposing-a-change | 16:33 |
| mgariepy | quick one.. when all galera node do seqno: -1 and no safe_to_bootstrap how do you pick the right node ? | 16:40 |
| mgariepy | i guess i take the first one since that's the one that gets the write :/ | 16:42 |
| noonedeadpunk | I close to never see seqno other then -1 | 16:44 |
| noonedeadpunk | and yeah - that's the principle I use as well :) | 16:44 |
| mgariepy | haha :) | 16:44 |
| mgariepy | i've seen some time that were not all seg_no: -1. | 16:45 |
| mgariepy | well i think it was more safe_to_bootstrap : 1 at the end. | 16:57 |
| *** dviroel|ruck|lunch is now known as dviroel|ruck | 17:05 | |
| spatel | noonedeadpunk This is how i integrate my glance with glusterFS for image service - https://satishdotpatel.github.io/openstack-ansible-glance-with-glusterfs/ | 17:09 |
| jrosser | the mount at the end looks a little unusual, am i misunderstanding that? | 17:16 |
| spatel | jrosser why unusal? | 17:27 |
| jrosser | now does it end up at /var/lib/glance in the container? | 17:28 |
| jrosser | *how | 17:28 |
| spatel | Yes /var/lib/gance is bind to -> /openstack/os-infra-1_glance_container-0a4a3789 | 17:28 |
| spatel | lxc.mount.entry = /openstack/os-infra-1_glance_container-0a4a3789 var/lib/glance/images none bind,create=dir 0 0 | 17:29 |
| spatel | lxc.mount.entry = /openstack/os-infra-1_glance_container-0a4a3789 var/lib/glance/cache none bind,create=dir 0 0 | 17:29 |
| spatel | its already there OSA does that lax.mount | 17:29 |
| jrosser | is that a default? | 17:31 |
| spatel | yes.. it came with osa deployment. | 17:32 |
| spatel | i didn't touch anything except just mount gfs and copy all data to new mount point | 17:33 |
| spatel | its working.. without any issue | 17:33 |
| jrosser | oh yes from here https://github.com/openstack/openstack-ansible/blob/master/inventory/group_vars/glance_all.yml#L19-L24 | 17:36 |
| spatel | I am thinking i can write playbook to do this integration if someone looking for this kind of solution. ofs openbstack-ansible-ops | 17:36 |
| jrosser | oh i see this https://github.com/openstack/openstack-ansible/blob/master/playbooks/os-glance-install.yml#L57 | 17:37 |
| spatel | :) | 17:38 |
| spatel | I found this solution cheaper and easy compare to swift deployment | 17:38 |
| noonedeadpunk | jrosser: I think we need align glance role with systemd_mount.... As since it's using it's own stanza that we convert later on, it makes things a bit weird imo... | 17:39 |
| jrosser | yes so those mounts only get made for the store being == file | 17:39 |
| spatel | ah! | 17:39 |
| noonedeadpunk | talkig about https://github.com/openstack/openstack-ansible-os_glance/blob/6bc829f2b2b9ab3bb192315398988f91e13ec764/tasks/glance_post_install.yml#L171-L176 | 17:40 |
| noonedeadpunk | specifically L173 | 17:40 |
| noonedeadpunk | so if you want to use cephfs, as example, you can't do this because what format is kind of hardcoded | 17:41 |
| jrosser | ah | 17:41 |
| jrosser | well line 188 doesnt help there either | 17:41 |
| noonedeadpunk | yum, it's not | 17:41 |
| noonedeadpunk | well, it's only naming though... | 17:42 |
| noonedeadpunk | but yes.... | 17:42 |
| noonedeadpunk | And last time I thought about that it was kind of hard to keep compatability | 17:42 |
| jrosser | yeah that needs a clean up | 17:42 |
| jrosser | looks like it's too specific in the glance role | 17:43 |
| jrosser | and maybe just deprecating glance_nfs_client totally and having a new format variable would be the answer | 17:43 |
| noonedeadpunk | well yes, by far I'd say that the only option | 17:44 |
| jrosser | seems we have lots of CI fail today | 17:45 |
| jrosser | i didnt get chance to look at it though | 17:45 |
| noonedeadpunk | on tempest plugins installation as git was failing checkout | 17:45 |
| noonedeadpunk | we don't have them as dependant projects I believe | 17:45 |
| noonedeadpunk | so they got cloned from opendev | 17:45 |
| noonedeadpunk | so infra thing I believe, but we need to fix this eventually as well | 17:46 |
| spatel | jrosser does my solution of glusterfs looks ok or any issue so i can take care right now before i go deploy everywhere | 17:49 |
| jrosser | i was just looking for how i made it take the services from the local git repos | 17:52 |
| jrosser | spatel: it looks fine - those mounts are only present for file backend which had me confused for a moment | 17:52 |
| spatel | cool! in my deployment i have backend is file everywhere :) | 17:54 |
| jrosser | noonedeadpunk: this would need extending to cover the tempest plugins https://opendev.org/openstack/openstack-ansible/src/branch/master/scripts/get-ansible-role-requirements.yml#L95-L112 | 17:55 |
| jrosser | which should be relatively straightforward as we already figure it out once | 17:55 |
| opendevreview | Danila Balagansky proposed openstack/openstack-ansible-os_nova master: Fix missing quote https://review.opendev.org/c/openstack/openstack-ansible-os_nova/+/834213 | 18:25 |
| *** dviroel|ruck is now known as dviroel|afk | 20:55 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!