| *** gcb has joined #openstack-interop | 01:23 | |
| *** markvoelker has quit IRC | 02:40 | |
| *** markvoelker has joined #openstack-interop | 03:44 | |
| *** pcaruana has joined #openstack-interop | 06:45 | |
| *** gcb has quit IRC | 08:12 | |
| *** gcb has joined #openstack-interop | 08:50 | |
| *** markvoelker has quit IRC | 09:25 | |
| *** cdent has joined #openstack-interop | 09:29 | |
| *** markvoelker has joined #openstack-interop | 10:26 | |
| *** markvoelker has quit IRC | 10:33 | |
| *** MarkBaker has joined #openstack-interop | 10:37 | |
| *** MarkBaker has quit IRC | 11:58 | |
| *** markvoelker has joined #openstack-interop | 13:23 | |
| *** markvoelker has quit IRC | 14:31 | |
| beisner_ | o/ dmellado | 14:48 |
|---|---|---|
| dmellado | hey beisner_ | 14:49 |
| dmellado | I wanted to check with tongli but he's not around | 14:49 |
| dmellado | maybe topol could now | 14:49 |
| topol | Hi dmellado, whats up? | 14:50 |
| dmellado | beisner_ was wondering if he could set http_proxy for the kubes master | 14:50 |
| dmellado | I kinda recall something about this, but could be for the previous' summit workload... | 14:50 |
| dmellado | so I was wondering if it would ring a bell for you | 14:51 |
| *** admcleod has joined #openstack-interop | 14:52 | |
| *** tongli has joined #openstack-interop | 14:53 | |
| topol | Tong said it should work but he didn't try it. He will login onto IRC now | 14:53 |
| *** markvoelker has joined #openstack-interop | 14:53 | |
| topol | ltongli^ | 14:54 |
| beisner_ | exporting http_proxy manually on the master resolves for us | 14:54 |
| topol | tongli^ | 14:54 |
| tongli | i am here | 14:54 |
| beisner_ | admcleod o/ | 14:55 |
| admcleod | beisner_: o\ | 14:56 |
| beisner_ | is there a systematic way to set http_proxy and no_proxy via the playbook or some such? | 14:56 |
| beisner_ | lol admcleod | 14:56 |
| tongli | @beisner_, notice the environment on each play in the site.yml file? | 14:57 |
| tongli | you should be able to use proxy_env to do whatever is needed. | 14:58 |
| tongli | but I have not tried it myself since I do not have a proxy_env. | 14:58 |
| admcleod | tongli: i see theres a var in group_vars/all.yaml, but that seems to be for the run, it doesnt seem to apply it to the instances that are launched? | 14:58 |
| tongli | no, see here https://github.com/openstack/interop-workloads/blob/master/workloads/ansible/shade/k8s/site.yml#L95 | 14:59 |
| tongli | http://docs.ansible.com/ansible/playbooks_environment.html | 14:59 |
| tongli | that should help. | 14:59 |
| tongli | I think if you just add proxy_env in your vars/xx.yml, that should work. | 15:00 |
| tongli | but I have never tried it myself. | 15:00 |
| admcleod | tongli: right.. yeah, thats just during the playbook run | 15:00 |
| dmellado | tongli: beisner_ | 15:01 |
| tongli | oh, you mean after it setup, you want to have a proxy? | 15:01 |
| dmellado | for the last iteration | 15:01 |
| admcleod | tongli: yep! | 15:01 |
| dmellado | we made it like here | 15:01 |
| dmellado | https://github.com/openstack/osops-tools-contrib/commit/f14020afc7731294fae29b8b0e26a3c83a141704 | 15:01 |
| dmellado | so maybe this could serve as an example | 15:01 |
| dmellado | althought I'd encourage to really, really run to try it | 15:01 |
| admcleod | dmellado: ive just run it using proxy_env, and the run works | 15:02 |
| dmellado | admcleod: on the ansible.cfg you mean? | 15:02 |
| admcleod | dmellado: problem is that the master then tries to pull down containers from gcr.io, but cant, because it doesnt have a proxy set | 15:02 |
| dmellado | admcleod I see | 15:03 |
| admcleod | dmellado: i didnt touch ansible.cfg... but group_vars/all.yaml has a proxy_env definition | 15:03 |
| dmellado | admcleod: I see, just exactly the same way as the example I put in there | 15:03 |
| admcleod | dmellado: right | 15:03 |
| dmellado | well, it wouldn't really hurt if you put a patch for it | 15:04 |
| admcleod | dmellado: ok :) | 15:04 |
| dmellado | that's just my 2 cents | 15:04 |
| dmellado | wdyt tongli topol ? | 15:04 |
| tongli | dmellado, not sure I am following. | 15:06 |
| dmellado | tongli: they'd like to add a patch for adding support for proxy just the way we had last time | 15:07 |
| tongli | we have environment in our playbook already. | 15:07 |
| tongli | we have that already. | 15:07 |
| dmellado | tongli: so you could set http_proxy in your env so the kube master would have a proxy? | 15:07 |
| dmellado | admcleod: beisner_ ^^ | 15:07 |
| tongli | I think what we have is to use the proxy at the run. | 15:07 |
| tongli | after that, how each node will access internet is outside of this workload. | 15:08 |
| beisner_ | http://paste.ubuntu.com/24511468/ | 15:08 |
| tongli | you can configure your node to use a proxy, right? | 15:08 |
| admcleod | tongli: correct | 15:08 |
| dmellado | but if the nodes are created at the workload | 15:09 |
| tongli | @admcleod, if you submit a patch, I can certainly take a look at. we are only few days from the demo, I am very hesitate to do any updates. | 15:09 |
| dmellado | they need to be configured | 15:09 |
| dmellado | yep, that was my idea too | 15:10 |
| tongli | if this is the issue that we found few weeks before, that will be addressed for sure. | 15:10 |
| tongli | one other thing that you can do in my view is to cache all the container images needed | 15:11 |
| tongli | cimage_repo was created for that purpose. | 15:11 |
| admcleod | tongli: ah.. | 15:11 |
| tongli | you can look at all k8s needed container images, and do docker pull, docker save. | 15:11 |
| tongli | which produce a tar, then tar gz it. | 15:12 |
| tongli | which produce a tar.gz file. | 15:12 |
| tongli | the workload already allows you to configure it in vars/xx.yml file. | 15:12 |
| tongli | you just need to manually produce that file and place it on a local http server. | 15:13 |
| tongli | point your workload in the vars/xx.yml file to that location. | 15:13 |
| tongli | you can run the workload absolutely isolated. | 15:13 |
| tongli | you can try that and using that also makes your workload runs very fast. | 15:13 |
| tongli | since it does not do the image pull. | 15:13 |
| tongli | but the problem with that is that you can not do phase 2 demo. | 15:14 |
| admcleod | tongli: ok, noted, thanks | 15:14 |
| tongli | let me know if you have other questions. | 15:15 |
| tongli | if you think it is really needed, please do a patch. | 15:15 |
| tongli | even if it is not merged, you can still run it with that patch. | 15:15 |
| dmellado | tongli: btw, related to the patch | 15:16 |
| *** galstrom_zzz is now known as galstrom | 15:16 | |
| dmellado | I'm fine running this on my own | 15:16 |
| dmellado | but it improves the speed quite a bit | 15:17 |
| dmellado | https://review.openstack.org/#/c/462508/ | 15:17 |
| tongli | but please with a patch though. | 15:17 |
| tongli | hmmm. you are changing the config file. during the run, we are dynamically creating this file run/runhost | 15:18 |
| tongli | without these things, I am not sure everything will still work. | 15:18 |
| tongli | the run/runhost is absolutely needed when we run the things in two separate phases. | 15:19 |
| tongli | @dmellado, please think about that as well. | 15:19 |
| dmellado | yep, well, my patch only affects to performance on the pipelining | 15:19 |
| dmellado | (and disables cows xD) | 15:20 |
| beisner_ | +1 for nowcows ;-) | 15:22 |
| beisner_ | s/nowcows/nocows/ | 15:22 |
| beisner_ | tongli dmellado many thx for input/assist | 15:24 |
| dmellado | np! ;) | 15:24 |
| tongli | sure | 15:25 |
| admcleod | dmellado: tongli yep thanks again | 15:31 |
| *** pcaruana has quit IRC | 16:40 | |
| *** catherineD has quit IRC | 17:24 | |
| *** catherineD has joined #openstack-interop | 17:30 | |
| *** cdent has quit IRC | 18:35 | |
| *** cdent has joined #openstack-interop | 18:47 | |
| *** catherineD|2 has joined #openstack-interop | 18:56 | |
| *** catherineD has quit IRC | 18:57 | |
| *** cdent has quit IRC | 19:54 | |
| *** cdent has joined #openstack-interop | 19:58 | |
| *** cdent has quit IRC | 20:47 | |
| *** pilgrimstack1 has quit IRC | 22:00 | |
| *** galstrom is now known as galstrom_zzz | 22:29 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!