*** sthussey has quit IRC | 00:31 | |
*** marst has quit IRC | 00:51 | |
*** marst has joined #airshipit | 01:05 | |
*** vamsis has joined #airshipit | 01:25 | |
*** vamsis has quit IRC | 01:29 | |
*** AhmadM has quit IRC | 02:10 | |
*** roman_g has quit IRC | 02:29 | |
*** jgu has joined #airshipit | 02:59 | |
*** vamsis has joined #airshipit | 03:14 | |
*** vamsis has quit IRC | 03:27 | |
*** vamsis has joined #airshipit | 03:43 | |
mattmceuen | marshallmargenau seaneagan https://storyboard.openstack.org/#!/story/2003014 for cfriesen_ 's ask above; can you take a look and give your thoughts? | 03:48 |
---|---|---|
*** cfriesen_ has quit IRC | 03:51 | |
mattmceuen | And please add in any additional pointers into the description for someone to implement it correctly - my armada-fu is not advanced enough to fill that in :) | 03:55 |
openstackgerrit | Van Hung Pham proposed openstack/airship-shipyard master: fix tox python3 overrides https://review.openstack.org/583455 | 04:42 |
*** jgu has quit IRC | 04:56 | |
*** adrianreza has quit IRC | 04:58 | |
*** d_dej has quit IRC | 05:09 | |
*** openstackgerrit has quit IRC | 05:10 | |
*** aagate has quit IRC | 05:16 | |
*** flaviosr has quit IRC | 05:16 | |
*** serverascode has quit IRC | 05:16 | |
*** fdegir has quit IRC | 05:16 | |
*** b-str has quit IRC | 05:16 | |
*** portdirect has quit IRC | 05:16 | |
*** flaviosr has joined #airshipit | 05:35 | |
*** aagate has joined #airshipit | 05:35 | |
*** serverascode has joined #airshipit | 05:35 | |
*** fdegir has joined #airshipit | 05:35 | |
*** b-str has joined #airshipit | 05:35 | |
*** portdirect has joined #airshipit | 05:35 | |
*** vamsis has quit IRC | 05:37 | |
*** vamsis has joined #airshipit | 05:40 | |
*** vamsis has quit IRC | 05:44 | |
*** vamsis has joined #airshipit | 06:05 | |
*** vamsis has quit IRC | 06:22 | |
*** cfriesen_ has joined #airshipit | 06:40 | |
*** vamsis has joined #airshipit | 06:44 | |
*** cfriesen_ has quit IRC | 06:45 | |
*** georgk has joined #airshipit | 06:49 | |
*** goutham has joined #airshipit | 07:14 | |
*** vamsis has quit IRC | 07:27 | |
*** vamsis has joined #airshipit | 08:17 | |
*** vamsis has quit IRC | 09:02 | |
*** vamsis has joined #airshipit | 09:07 | |
*** vamsis has quit IRC | 09:08 | |
*** roman_g has joined #airshipit | 09:28 | |
*** vamsis has joined #airshipit | 10:22 | |
*** vamsis has quit IRC | 10:26 | |
*** eanylin has joined #airshipit | 10:37 | |
*** vamsis has joined #airshipit | 11:39 | |
mattmceuen | Actually cfriesen_ - have you checked out `armada tiller --releases` for listing releases? I think it's just armada-managed manifest listing that might be missing - let me know if that's something you need or if releases fits the bill | 12:08 |
portdirect | If we were going to list armada manifest managed releases, we probably need armada to start storing state somewhere? | 12:12 |
roman_g | cfriesen_ is not here atm. | 12:26 |
roman_g | mattmceuen: is it a bug or feature in irccloud that you and others often reply to persons who are not currently on a channel? ) | 12:27 |
portdirect | roman_g: it's pretty common in openstack | 12:29 |
portdirect | As the channels are logged | 12:29 |
portdirect | http://eavesdrop.openstack.org/irclogs/%23airshipit/%23airshipit.2018-07-18.log.html | 12:29 |
roman_g | Got it. But I'm not sure that one would go to logs and search his/her nickname for mentions. | 12:30 |
*** openstackgerrit has joined #airshipit | 12:30 | |
openstackgerrit | Smruti Soumitra Khuntia proposed openstack/airship-shipyard master: A new Shipyard status API and CLI https://review.openstack.org/582156 | 12:30 |
roman_g | portdirect: re: armada to start storing state somewhere - in annotations? | 12:31 |
portdirect | That's one possibility | 12:32 |
portdirect | Or using a crd/configmap | 12:33 |
portdirect | Theres lots of ways to skin that cat, but this has always been the hurdle that has to be overcome to start getting these features added | 12:34 |
portdirect | An alternate would be to have an 'armada get -f template.yaml' type command | 12:35 |
portdirect | That would be somewhat analogous to similar kubectl commands | 12:36 |
openstackgerrit | Smruti Soumitra Khuntia proposed openstack/airship-shipyard master: A new Shipyard status API and CLI https://review.openstack.org/582156 | 12:42 |
openstackgerrit | Smruti Soumitra Khuntia proposed openstack/airship-shipyard master: A new Shipyard status API and CLI https://review.openstack.org/582156 | 12:44 |
*** sthussey has joined #airshipit | 13:47 | |
*** eanylin has quit IRC | 13:55 | |
*** jgu has joined #airshipit | 14:02 | |
*** jgu has quit IRC | 14:05 | |
*** jgu has joined #airshipit | 14:06 | |
*** vrv_ has joined #airshipit | 14:14 | |
*** vamsis has quit IRC | 14:26 | |
*** vamsis has joined #airshipit | 14:28 | |
*** aaronsheffield has joined #airshipit | 14:33 | |
*** vamsis has quit IRC | 14:33 | |
seaneagan | mattmceuen: I added my thoughts to https://storyboard.openstack.org/#!/story/2003014 | 14:40 |
seaneagan | re: cfriesen_'s ask about rollback of whole manifest, that has been discussed, but would similarly require armada to track manifest applications and their releases | 14:45 |
sthussey | what does it mean to rollback a manifest? | 14:50 |
portdirect | sthussey: https://github.com/kubernetes/helm/blob/master/docs/helm/helm_rollback.md | 14:50 |
portdirect | though i think in an airship world, it may make more sense to roll-forward to the previous config... | 14:51 |
sthussey | manifest 1 = [chart A(overrides X)], manifest 2 = [chart A(overrides X')]. Apply manifest 1, apply manifest 2, rollback manifest 1 | 14:51 |
sthussey | what is the end result? | 14:51 |
*** vamsis has joined #airshipit | 14:52 | |
portdirect | i dont think thats possible | 14:52 |
portdirect | you need to have a revision history in tiller to roll back to | 14:52 |
portdirect | but for this to have real world value therea a lot that needs to be acounted for that helm cannot do | 14:53 |
portdirect | eg potential db migrations etc | 14:53 |
*** goutham has quit IRC | 14:55 | |
*** cfriesen_ has joined #airshipit | 14:55 | |
sthussey | Mostly my point - I'm not sure that since rollback is basically a proxy for helm/tiller functionality today and the manifest construct doesn't exist in helm that 'rollback a manifest' may not have a clear meaning | 15:20 |
sthussey | And just issuing a helm rollback for every chart resulting from a manifest may not be the right path | 15:21 |
*** jgu has quit IRC | 15:27 | |
*** jgu has joined #airshipit | 15:27 | |
*** jgu has quit IRC | 15:31 | |
*** vamsis has quit IRC | 15:57 | |
*** georgk has quit IRC | 15:59 | |
*** vamsis has joined #airshipit | 16:09 | |
cfriesen_ | sthussey: as I understand it, armada is supposed to manage a set of helm charts with interdependencies and triggers and such. It seems odd to me that you can "apply" a manifest (which will either do an install or upgrade of the underlying charts) but you can't do the equivalent of helm's list/status/inpect/rollback commands. | 16:28 |
cfriesen_ | if I do an "apply" of a new manifest and it fails, how do I get back to the previously-working state? Without a manifest rollback I need to keep the previous version of the manifest around in external tooling so I can "apply" it again. | 16:29 |
cfriesen_ | that's doable, but it's different from helm (and armada bills itself as a helm client replacement) | 16:30 |
openstackgerrit | Mark Burnett proposed openstack/airship-promenade master: Enable etcd helm test to run on non-ready nodes https://review.openstack.org/583661 | 16:32 |
openstackgerrit | Mark Burnett proposed openstack/airship-promenade master: Tolerate unready endpoints for apiserver service https://review.openstack.org/583662 | 16:32 |
openstackgerrit | Mark Burnett proposed openstack/airship-promenade master: Fix broken validatedesign endpoint https://review.openstack.org/583663 | 16:32 |
openstackgerrit | Mark Burnett proposed openstack/airship-promenade master: Add liveness probe to kube-proxy https://review.openstack.org/583664 | 16:32 |
cfriesen_ | similarly, it seems like I'd need to have external tooling to even keep track of what manifests have been applied since there's no way to list them. | 16:35 |
openstackgerrit | Hari proposed openstack/airship-shipyard master: Block site_update if there is no host in MaaS/Dryd https://review.openstack.org/583665 | 16:46 |
*** vrv_ has quit IRC | 17:33 | |
sthussey | no disagreement. will just need to clearly spec out what rolling back a manifest means | 17:40 |
*** vamsis has quit IRC | 18:21 | |
*** felipemonteiro_ has joined #airshipit | 19:24 | |
*** felipemonteiro_ is now known as felipemonteiro | 19:24 | |
*** felipemonteiro_ has joined #airshipit | 19:30 | |
*** felipemonteiro has quit IRC | 19:33 | |
openstackgerrit | Ahmad Mahmoudi proposed openstack/airship-in-a-bottle master: Updated dev-minimal to deploy behind proxy https://review.openstack.org/582665 | 19:33 |
*** felipemonteiro_ has quit IRC | 19:50 | |
*** felipemonteiro_ has joined #airshipit | 19:50 | |
*** jgu has joined #airshipit | 20:01 | |
*** jgu has quit IRC | 20:02 | |
cfriesen_ | I tried running the armada deployment scripts in openstack-helm/tools/deployment/armada and I notice that it didn't set up any kubernetes resources (ie requests/limits for cpu/memory). Is this intentional? How is it happening given that the openstack-helm charts have default resources in them? (Or is armada not using those helm charts?) | 20:09 |
sthussey | @cfriesen_ did you override this default? https://github.com/openstack/openstack-helm/blob/803faedac0866535e84d08bef2ee22f67c447ff5/heat/values.yaml#L1016 | 20:12 |
cfriesen_ | sthussey: ah, that's probably it. thanks. | 20:17 |
*** vamsis has joined #airshipit | 20:21 | |
*** jgu has joined #airshipit | 20:26 | |
*** vamsis has quit IRC | 20:26 | |
*** felipemonteiro__ has joined #airshipit | 20:48 | |
*** aaronsheffield has quit IRC | 20:49 | |
*** felipemonteiro_ has quit IRC | 20:52 | |
openstackgerrit | Marshall Margenau proposed openstack/airship-armada master: Wait for jobs to complete https://review.openstack.org/583738 | 20:55 |
cfriesen_ | Starting with my armada-installed openstack-helm, as a test I ran "armada apply /tmp/armada-osh.yaml --set chart:mariadb:values.pod.resources.enabled=true" This resulted in a new helm release for osh-mariadb with kubernetes resources reserved. (As expected.) | 20:59 |
cfriesen_ | However, it also resulted in new helm releases for osh-memcached, osh-openstack-ceph-config, and osh-openstack-ingress-controller, and the overall armada call seems to have hung. | 20:59 |
*** jgu has quit IRC | 21:00 | |
cfriesen_ | what's the bug report procedure? storyboard story? | 21:00 |
sthussey | @mattmceuen ^ | 21:02 |
openstackgerrit | Marshall Margenau proposed openstack/airship-armada master: Removing dead code. https://review.openstack.org/583740 | 21:04 |
*** felipemonteiro__ has quit IRC | 21:07 | |
*** felipemonteiro__ has joined #airshipit | 21:07 | |
marshallmargenau | @cfriesen_ when armada looks hung, it's usually a result of it waiting on something that wasn't defined quite right in your charts | 21:11 |
marshallmargenau | i believe it should wait 15 minutes by default unless you specified something different | 21:12 |
marshallmargenau | the logs should also show you what namespace and label selectors (including none, which is the most common cause) that it's currently waiting to resolve | 21:12 |
marshallmargenau | will defer to @mattmceuen or @mark-burnett on bug report procedure | 21:13 |
cfriesen_ | marshallmargenau: are there logs other than stdout? the last stdout line is just "2018-07-18 20:26:48.465 24313 INFO armada.handlers.tiller [-] Deleting pod osh-keystone-rabbitmq-test in namespace: openstack" | 21:16 |
cfriesen_ | marshallmargenau: the manifest/charts are untouched from when I initially ran the armada scripts to bring up openstack-helm, the only change should have been to enable resources for mariadb. | 21:17 |
cfriesen_ | and it's been hung for almost an hour | 21:18 |
*** felipemonteiro__ has quit IRC | 21:28 | |
cfriesen_ | doing ctrl-C and then rerunning the same command seems to now allow it to get further. | 21:28 |
cfriesen_ | looking at the helm charts from the two releases, it looks like the output of "helm get <release> --revision <X>" for the two revisions is functionally identical but reordered slightly. I suspect this caused something to think it had changed when it actually hadn't. | 22:15 |
cfriesen_ | the above command for the osh-keystone helm release results in a yaml file with 63 docs, and of these six were reordered between the two releases. | 22:19 |
cfriesen_ | For osh-memcached though only the revision number and the release timestamp were different. | 22:24 |
openstackgerrit | Bryan Strassner proposed openstack/airship-shipyard master: [WIP] Add redeploy_server processing https://review.openstack.org/583793 | 23:02 |
*** vamsis has joined #airshipit | 23:14 | |
*** vamsis has quit IRC | 23:18 | |
openstackgerrit | Felipe Monteiro proposed openstack/airship-pegleg master: Add git and branch revision support to pegleg https://review.openstack.org/582652 | 23:34 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!