[09:35] niemeyer: https://bugs.launchpad.net/ensemble/+bug/771348 [09:35] <_mup_> Bug #771348: status barfs when checking newly deployed service units (no state) < https://launchpad.net/bugs/771348 > [09:37] bcsaller: Cheers [09:40] hazmat, dublin is fun + productive [09:40] A new screencast springs into existence → http://www.youtube.com/user/ubuntucloud#p/a/u/0/AMHcy63wRL0 [09:40] "Ensemble deploy and scale cloud apps with ease" [09:41] kim0, sounds great! need to get my headphones out [09:41] if you spot a horrible mistake .. let me know .. can still fix it [11:50] <_mup_> ensemble/expose-status r263 committed by jim.baker@canonical.com [11:50] <_mup_> Tests to verify that open-ports is not set in status unless service is exposed [11:51] <_mup_> ensemble/expose-status r264 committed by jim.baker@canonical.com [11:51] <_mup_> Do not set open-ports unless service is exposed [12:23] <_mup_> Bug #802931 was filed: Modify provisioning agent so that the machine opens/closes ports < https://launchpad.net/bugs/802931 > [12:57] Does anyone agree "ensemble shutdown" should have a -y option ? [13:35] can ensemble formula be interactive? [13:36] meaning, can it ask for some user input before for information required when packages are installed [14:10] ivoks: afaik not really .. but we'll get that soon with config-set (passing variables to formulas) [14:11] devs correct me if I'm wrong [14:11] ok, that's good enough [14:12] how does expose/unexpose work? [14:13] even if i don't expose, all the services from my formula are available from outside [14:13] ivoks: I believe expose is for opening up ports using ec2 security groups (current provider) [14:13] but today .. I believe all ports are open by default [14:13] ok [14:14] i'm working on mail-stack formula [14:14] i got it working fine, with exception of not being able to define relayhost :) [14:15] Ben is working hard to add passing params [14:15] should hopefully land soon [14:15] what i'd like to do more is to separate postfix to one instance and dovecot to the other [14:16] ivoks: btw https://ensemble.ubuntu.com/docs/ has more info on exposing :) [14:16] yeah, i know [14:16] :) [14:20] ivoks, atm things are open, but in future, when the ports stuff land it will be closed to public by default but open to internal traffic (till we adjust the ports work to take into account relations) [14:21] ok [14:21] ivoks, re how ports work.. the formula hooks can designate which hooks they want to open and close via an ensemble cli api (open-port, close-port).. the ensemble user/admin will need to expose the service for these hook defined settings to take effect [14:22] ok [14:24] underneath the hood their using ec2 security groups as kim0 pointed out [14:24] kim0, are you in dublin this week? [14:25] Yes [14:25] you are not right [14:25] kim0, nope.. just wondering where the dev team is hanging out at [14:26] kim0, i did a mongodb conference yesterday.. and did a quick demo of the mongodb formula i've been working on [14:26] but it also needs config-set for the sharding configuration [14:26] woohoo [14:26] hazmat: share that formula man :) [14:27] * hazmat looks up the how to share with principia docs [14:27] niemeyer, hallo! [14:29] destroy-service doesn't really do anything yet? [14:29] ivoks, it does.. it removes relations, destroys units, it should wipe the service from the environment [14:29] hm [14:29] everything is still running on my instance :) [14:30] hmm.. [14:30] ivoks, have you implemented the stop hook? [14:30] there are no relations, but dovecot, postfix and amavis are still running [14:31] yes; it might be wrong [14:31] service dovecot stop [14:31] same for postfix and amavis [14:31] * hazmat checks the implementation [14:32] SpamapS, how's the reception of lxc-provider at the rally? [14:39] ivoks, so it doesn't appear that destroying a service will actually inform the unit it should suicide, it just tells the machine agent to kill the unit agent [14:39] ivoks, so in the destroy-service case with the service processes still running, do you see any unit agent processes in the process listing? [14:40] so the relation-broken hooks are invoked, but not afaics the stop hook [14:40] i see unit agent [14:40] /usr/bin/python -m ensemble.agents.unit -n --pidfile /var/lib/ensem.... [14:41] hmm [14:41] it's not listed in ensemble status [14:42] ah, don't worry about it [14:42] i have other things to solve first [14:42] hazmat: Hey man! [14:42] hazmat: Missing you here [14:42] ivoks, i'll keep digging into it [14:42] like 'restart service after 'install'' [14:42] niemeyer, how's it going so far [14:42] hazmat: Fantastic [14:42] niemeyer, awesome.. any highlights so far? or just a good productive vibe going around? [14:42] niemeyer: i bet it's raining all the time :) [14:43] ivoks: :-) [14:43] hazmat: Quite a few.. still have to digest [14:43] hazmat: What are you working on right now? [14:43] i think its rained every day at wimbledon but like two.. out of 10 [14:43] hazmat: and how much time would it take for you to get rid of our pre-baked images? [14:43] niemeyer, atm digging into destroy-service [14:43] niemeyer, maybe 2.5 hrs [14:43] hazmat: Awesome! [14:43] hazmat: Can you push that? [14:44] niemeyer, sounds good, i'll put an issue to have a look at destroy-service later [14:44] hazmat: Yeah, destroy-service actually looks important, but I wonder if it should be postponed and solve LXC first [14:45] hazmat: destroy-service will likely have a different story in that light [14:45] hazmat: also there is some hinting at more review on my branch which would be helpful [14:47] niemeyer, not sure that its significantly different, in light of lxc.. perhaps.. the machine agent does a destroy on the unit deployment, and the relations are broken first.. i guess it depends on how we structure the agents for the container, if a machine agent lives in each container in the provider case, or if we just have a singleton machine agent on the physical machine creating containers (which would translate well to pushing the work to ec2) [14:47] bcsaller, indeed i do need to follow up on that [14:47] hazmat: Yeah, maybe it's not that different indeed [14:48] hazmat: What about the security stuff? [14:48] hazmat: Would be good to get some action happening on that front sooner rather than later [14:49] <_mup_> Bug #802995 was filed: Destroy service should invoke unit's stop hook, verify/investigate this is true < https://launchpad.net/bugs/802995 > [14:50] niemeyer, i've got some comments on the review, i'll try to kick that back out today.. so reviews and image work on my plate for today [14:50] hazmat: Cool.. I'm just concerned we're not getting any action for a while [14:51] hazmat: We can continue evolving the spec and the overall design, but would be good to define what are the first few steps and get these going [14:52] hazmat: note that this was reproduced on my custom formula (maybe the formula is broken) [14:53] niemeyer, the mongodb conference yesterday was fun.. their roadmap has some cool features.. the aggregation stuff (map/reduce replacement) is really nice [14:54] hazmat: It's awesome indeed.. really looking forward to it [14:58] off to dog-walk, back in a few [14:58] hazmat: Have fun [16:55] hazmat: re the lxc provider, its a bit rough getting going, but m_3 was able to get it going [16:55] SpamapS nice, you doubled the user population ;-) [16:55] <_mup_> ensemble/expose-provision-machines r264 committed by jim.baker@canonical.com [16:55] <_mup_> Initial commit [16:56] hazmat: I really want to get all of the agents running under upstart, that would be cool. [16:58] SpamapS, i had a look after we discussed last month, it looked pretty straightforward [16:58] i'll see if i can get in this week if no else gets to it [17:03] <_mup_> Bug #803042 was filed: The "bootstrap node" needs high availability < https://launchpad.net/bugs/803042 > [17:09] i could help with that :) [17:11] Yos [17:11] ivoks: Why aren't you here? :) [17:11] don't know :) [17:24] <_mup_> ensemble/sans-ami r264 committed by kapil.thangavelu@canonical.com [17:24] <_mup_> use standard ubuntu amis [17:24] <_mup_> ensemble/sans-ami r265 committed by kapil.thangavelu@canonical.com [17:24] <_mup_> add missing natty release image data for tests === koolhead11 is now known as koolhead11|afk [17:56] <_mup_> ensemble/expose-provision-machines r265 committed by jim.baker@canonical.com [17:56] <_mup_> Now loops [18:00] <_mup_> ensemble/expose-provision-machines r266 committed by jim.baker@canonical.com [18:00] <_mup_> Removed debugging sleeps from new test === daker is now known as daker_ [19:01] <_mup_> ensemble/sans-ami r266 committed by kapil.thangavelu@canonical.com [19:01] <_mup_> allow image specifications to select an ami, these specs are not passed to user data. [20:07] <_mup_> ensemble/sans-ami r267 committed by kapil.thangavelu@canonical.com [20:07] <_mup_> use the ensemble ppa for software installation by default, but also support branch installations. [20:48] <_mup_> ensemble/sans-ami r268 committed by kapil.thangavelu@canonical.com [20:48] <_mup_> tweak install commands to include state and log dir creation [21:13] <_mup_> ensemble/sans-ami r269 committed by kapil.thangavelu@canonical.com [21:13] <_mup_> correct name for txzookeeper, and normalize checkout dirname [22:39] * niemeyer breaths and heads to some reviewing.. [22:40] A bit sleepy.. may not last long. :)