[00:18] Okay, some great code reviewed, and now it's really time for bed. [00:18] Night all [09:51] idea: does it make sense to use micro for example? [10:17] ivoks: You mean the micro instance type? [10:17] yes [10:17] for bootstrap [10:17] ivoks: It's too small [10:17] ivoks: Well.. maybe for bootstrap it's not [10:17] ivoks: But in general it's too slugish [10:17] sluggish [10:18] i know [10:18] ivoks: We actually used it across the board until very recently [10:18] but for learning and playing with ensemble... [10:18] ivoks: But we had complaints about it being extremely unpredictable in terms of performance [10:18] amazone gives two micro instances for free for 1 year [10:19] ivoks: So we gave up on the idea to avoid burning the project image just because the instance is unreliably slow [10:19] for someone that wants to try and play with ensemble (no production), those would be enough [10:19] ok [10:20] ivoks: Yeah, it's a bit of a pity, but there's a reason why they give it for free :-) [10:23] i know :) === daker_ is now known as daker [11:25] <_mup_> ensemble/expose-provision-machines r267 committed by jim.baker@canonical.com [11:25] <_mup_> Add test re assigning a machine to a service unit with opened ports, opens ports on that machine === koolhead11|afk is now known as koolhead11 [14:55] * hazmat waits for the coffee to kick in [16:13] hazmat: ping [16:14] niemeyer, pong [16:14] hazmat: Yo [16:14] hazmat: All good there? [16:14] just chatting with bcsaller about the config-set branch and orchestra [16:14] niemeyer, all good.. wimbledon on in the background [16:14] niemeyer, working on security doc, and upstart today [16:14] hazmat: Ah, awesome [16:15] niemeyer, ready for the meeting anytime [16:15] hazmat: How's the image stuff? [16:15] niemeyer, in review [16:15] hazmat++ [16:15] hazmat: I expect to have a draft doc of the co-located services proposal out soon too, so you can hear our thinking on that [16:15] niemeyer, looking good, start time for the bootstrap node is a little longer, but the rest seems quite workable [16:15] hazmat: Can I trow something else on your table without you falling down? [16:15] throw even [16:15] niemeyer, shoot, we'll see if it sticks ;-) [16:16] hazmat: Juan had difficulties with peer relations [16:16] hazmat: It's my understanding that they were supposed to work already, is that right? [16:16] niemeyer, hmm.. how so.. i've been using them a little for the mongodb stuff, but we don't have many formulas that use them [16:16] hazmat: the peer events didn't seem to fire, we haven't looked into it yet [16:16] actually any [16:16] hmm [16:16] hazmat: Hah, sweet [16:16] niemeyer, i'll dig into that [16:16] hazmat: So it's likely something minor [16:16] hazmat: Let me send you his formula [16:16] niemeyer, which formula .. thanks [16:17] i can prioritize that, its probably a minor one [16:18] hazmat: Should be in your inbox [16:20] niemeyer, not seeing it yet [16:21] niemeyer, got it [16:21] hazmat: Sweet [16:22] niemeyer, no attachment though [16:22] hazmat: Meeting in 1:40.. will try to dial you in a little earlier [16:22] hazmat: The branch is there [16:22] niemeyer, ah right.. regular phone or skype? [16:22] hazmat: What phone number should I dial? [16:22] niemeyer, in priv msg [16:23] hazmat: Thanks === bcsaller1 is now known as bcsaller [16:39] negronjl, so the problem with the tomcat formula is that its using relative addressing to files [16:39] 2011-06-29 11:38:28,105 unit:tomcat6/1: hook.output ERROR: sed: can't read /usr/lib/ensemble/txzookeeper/../templates/tomcat-server-xml-top.tmpl: No such file or directory [16:39] hth [16:39] that output is from the debug-log which is very useful for debugging problems.. as is the debug-hooks cmd [16:40] negronjl, there's a branch in review to ensure that formula hooks are always executed from the formula dir [16:40] but it hasn't landed, atm the hook PWD isn't something that can be relied on by hooks [16:40] and addressing paths from there should be absolute [16:40] negronjl, i guess in this case your trying to get to some formula included config? [16:41] negronjl, which will become more easier once bug 761015 is fixed [16:41] <_mup_> Bug #761015: FORMULA_DIR should be available to hooks < https://launchpad.net/bugs/761015 > [16:41] * hazmat grabs some food === koolhead11 is now known as koolhead11|afk [18:06] hazmat: ping? [18:06] niemeyer, pong [18:06] hazmat: Ring.. ring.. ring.. [18:06] hazmat: Ring.. ring.. ring.. [18:06] niemeyer, no ring here [18:06] niemeyer, phone or skype? [18:06] hazmat: We just talked to your answering machine [18:07] niemeyer, hmm.. my phone is low let me plug-in and see if works better [18:07] hazmat: Again [18:19] niemeyer, bcsaller, jimbaker, anyone.. i dropped.. pls callback [18:20] hazmat: Calling [18:20] niemeyer, thanks [19:05] niemeyer, bcsaller, jimbaker, anyone.. i dropped again.. pls callback [19:05] hazmat Calling [19:16] niemeyer, bcsaller, jimbaker, anyone.. argh! i dropped again.. pls callback [19:17] * hazmat kicks his phone company where the light don't shine [19:19] niemeyer, bcsaller, jimbaker, anyone.. argh! i dropped again.. it might be time to give up.. [20:03] niemeyer, i've been curious while looking at zeromq go bindings, how the general problem of thread local resources is handled in go, since there isn't any easy way of distinguishing thread from a goroutine (ie their mapped m to n to threads) [20:04] seems like its also relevant in the context of gui apps where typically only one thread is managing interacting with the gui run loop [20:06] the only thing variant i've seen for this seems a bit like overkill (runtime.LockOSThread etc) [20:07] hazmat: We have exactly the same problem with zk [20:07] hazmat: It's handled by giving it its own thread [20:07] hazmat: But the zeromq Go author wasn't very clever at that [20:08] hazmat: Which means it will blow up on people's face [20:08] indeed, it looks experimental at best [20:08] interestingly the erlang for (zeromq) version does much the same (separate thread allocation) [20:08] niemeyer, thanks, i'll take a look at the gozk impl for ideas [20:09] hazmat: Haven't seen that, but it's the only option that makes sense.. goroutines with free threading is a feature which we shouldn't undermine === daker is now known as daker_