Wednesday, 2017-10-11

*** rbudden has quit IRC02:02
*** jmlowe has joined #scientific-wg03:24
*** jmlowe has quit IRC04:33
*** blair_ has joined #scientific-wg05:57
blair_anyone seen guest crashes correlated with KVM rdmsr errors before ?07:23
*** blair_ has quit IRC07:31
*** priteau has joined #scientific-wg09:20
-openstackstatus- NOTICE: The CI system will be offline starting at 11:00 UTC (in just under an hour) for Zuul v3 rollout: http://lists.openstack.org/pipermail/openstack-dev/2017-October/123337.html10:09
*** b1airo has joined #scientific-wg10:29
b1airohi all, we are hereby known as the Scientific SIG (https://wiki.openstack.org/wiki/Scientific_SIG)10:55
*** martial has joined #scientific-wg10:56
*** masuberu has joined #scientific-wg11:05
*** masber has quit IRC11:07
*** rbudden has joined #scientific-wg11:28
*** khappone has joined #scientific-wg12:01
khapponeHello *12:01
b1airohi12:01
martialhello again :)12:02
khapponeWe standardize the login names for all our images. We do turn on automatic updates on all images (I don't know the upstream standard though), which is kind of important. However I'm not sure if it's important enough to deviate from the standard12:02
khapponeSome sites made the images use local NTP servers12:02
b1airoso is your new approach to not standardise login names?12:03
b1airohere are our image build recipes: https://github.com/NeCTAR-RC/nectar-images12:04
khapponeBasically we'd take what ubuntu, centos et.al. offer and use them directly12:04
khapponeWe'd update our documentation, have a few mails and a small announcement capaign to our users.12:04
b1airoyeah that makes sense, we try to be hands off - things like the usernames are an informal standard across clouds, e.g., most experienced users know and rely on Ubuntu images using "ubuntu" etc12:05
khapponeOne of the point is also to make possible future cloud federation and collaboration easier12:05
b1airoyou can see the base of what we change here: https://github.com/NeCTAR-RC/nectar-images/blob/master/ansible/playbook.yml12:05
khapponeYeah, I'm still torn if this is a good idea. I'd really love to somewhat standardize on this, but I also think there are relevant tweaks12:06
b1airoyep, it's tough - some of this is strictly necessary for us, e.g., installing fail2ban12:07
b1airoour cloud instances are on the Internet by default, and even though secgroups default to being closed, many users still just open 0.0.0.0/012:07
khapponeAnd you feel like key only SSH isn't safe enugh?12:08
b1airoand of course we have guidelines telling them not to set passwords etc, but half the time they have no idea what they are doing12:08
khapponeAh, we have solved this by never knowing what they're doing and when something goes wrong we'll shut their VM(s) down :)12:09
b1airoha! i wish my life were that simple12:09
khapponeBut I think it's a compromise between technical controls and documentation12:10
khapponeand defaults12:10
martial:)12:10
khapponeI don't think anybody uses the recommendations I wrote yet (we don't yet), so I don't know if they're reasonable. But we'll try to figure it out. And it might be that even the base images would have some agreed-upon tweaks in them12:11
b1airoi think it would be useful to extend your work to try and find where a sensible cut-off point is for customisation, i suspect a blanket "no customisation" rule will not be practical, but if we can come up with reasoned guidelines about what should not be touched then that would be useful to avoid interoperability issues for users12:11
b1airothis would have made a really great Forum topic12:12
b1airoperhaps in 6 months if you can be there!12:12
khapponeIf we do that, I'd also like to have the same standard customizations. Partly to make the images and tools reusable, and secondly to be able to build on that and share more complex appliances12:13
khapponeI'll try to be in Vancouver12:13
martialmaybe a customization interface?12:13
martialansible playbook calling other ansible playbooks12:13
khapponeI'm sure the compromize won't be what exactly everybody wants, but if the benefits outweigh the comromise (which I think could easily be done), we could get someting to work12:14
b1airokhappone, do you have any examples of times that customisations have caused problems for end-users/consumers ?12:14
khapponeb1airo: yes, we can't directly steal your images, so it's extra work for us, and we can't focus on what the users really need :P12:15
b1airoit may be useful to list some of those out to use as sanity tests12:15
b1airohaha12:15
khapponeFinal goal: The only images we build are something that's not out there, and nobody else has to build that.12:16
b1airomaybe you could directly steal our image code though... just a few new group_vars :-)12:16
khapponeThat's an option, a less-nice one, but an option. Sharing tools is also beneficial. But I WANT MORE.12:17
khappone:)12:17
khapponeAnyway, I really need to run now. I'm glad there was discussion around this topic. I shall try work on it on our part and see what problems I find.12:18
martialsounds very cool, thanks for sharing the link12:18
b1airothanks for staying to chat khappone !12:18
b1airotime for me to head to bed as well12:18
martialgood night then b1airo12:19
martialgood bye khappone12:20
*** martial has quit IRC12:20
*** priteau has quit IRC12:30
-openstackstatus- NOTICE: Due to unrelated emergencies, the Zuul v3 rollout has not started yet; stay tuned for further updates13:06
khapponeb1airo: when you're back. May I add the link to your image build tools to the document?13:32
*** priteau has joined #scientific-wg13:35
*** jmlowe has joined #scientific-wg13:55
*** rbudden has quit IRC15:36
*** rbudden has joined #scientific-wg15:38
*** jmlowe has quit IRC16:47
*** jmlowe has joined #scientific-wg18:18
*** masuberu has quit IRC19:01
*** jmlowe has quit IRC19:30
*** jmlowe has joined #scientific-wg19:48
*** priteau has quit IRC20:38
*** jmlowe has quit IRC20:38
*** jmlowe has joined #scientific-wg20:57
*** b1airo has quit IRC21:06
*** jmlowe has quit IRC21:12
*** jmlowe has joined #scientific-wg21:25
*** jmlowe has quit IRC21:47
*** jmlowe has joined #scientific-wg21:53
*** b1airo has joined #scientific-wg21:57
*** b1airo has quit IRC22:05
*** priteau has joined #scientific-wg22:39
*** priteau has quit IRC22:44
*** jmlowe has quit IRC22:53
*** priteau has joined #scientific-wg23:01
*** priteau has quit IRC23:06
*** masber has joined #scientific-wg23:14
*** b1airo has joined #scientific-wg23:39

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!