/srv/irclogs.ubuntu.com/2012/04/02/#juju-dev.txt

=== TheMue_ is now known as TheMue
=== bigjools-afk is now known as bigjools
=== asavu_ is now known as asavu
niemeyerHello there!12:58
TheMueniemeyer: moin13:00
niemeyermthaddon: ping13:31
mthaddonniemeyer: hi hi13:32
niemeyermthaddon: Yo13:32
niemeyermthaddon: How're things going in the store?13:32
mthaddonniemeyer: I posted an update to the RT with the things we still need to look at (mostly around deployments and mongo authentication)13:32
niemeyermthaddon: I've seen it13:33
niemeyermthaddon: re. auth, you can actually put auth data in the command line arg already13:33
mthaddonah cool - how do we do that? and will that mean the auth info shows up in "ps"?13:33
mthaddonniemeyer: ^13:37
niemeyermthaddon: Hmm, yeah, potentially13:37
niemeyermthaddon: You can use a URL such as mongodb://foo:bar@address/juju13:38
mthaddonniemeyer: can we pull from a config file instead? otherwise it kind of defeats the point of having any auth info :)13:38
niemeyermthaddon: Yeah, of course13:39
mthaddongreat13:39
niemeyermthaddon: Will have that ready to go in a moment.. have your read about protecting a db?13:39
niemeyermthaddon: You can workaround it, btw.. even if it's boring.. but I'm happy to make that saner for you13:40
mthaddonniemeyer: I'm happy to work around it if there's a good way of doing that - what is that?13:40
niemeyermthaddon: Just testing now to see if I'm not speaking crack13:47
mthaddonk, thx13:48
niemeyermthaddon: Nevermind, I am crazy13:48
* mthaddon not minding13:48
niemeyermthaddon: Either way, I'd be unhappy with such a hack.. I'll have a more proper fix13:48
niemeyermthaddon: Btw, meanwhile, have you read about the auth stuff on dbs?14:16
mthaddonniemeyer: the docs don't seem very in depth, but I have read http://www.mongodb.org/display/DOCS/Security+and+Authentication14:26
mthaddonwhy on earth you'd have an "auth = True/False" and a "noauth = True/False" option in the configs seems beyond me...14:27
niemeyermthaddon: Cool, that should be enough14:30
niemeyermthaddon: You mean to have both, or to have it at all?14:30
mthaddonniemeyer: having both - can only lead to confusion14:31
niemeyermthaddon: Yeah, I'd guess it's something to do with the overriding of defaults14:31
niemeyermthaddon: But, I've never used both, so have no idea really14:31
TheMueso, leaving today for my Go talk at the GTUG14:32
niemeyerTheMue: Ah, super, good luck there, and please let us know how it goes14:33
TheMueniemeyer: sure, and I'll add the presentation to slideshare later14:33
niemeyerTheMue: Sweet14:35
niemeyerrobbiew: Is your wifi going bad as well?14:35
hazmatjimbaker, ping14:48
jimbakerhazmat, hi14:48
hazmathi jimbaker just going through rel hook context, looks good, but one question i had was if there was testing around rel mutation during the execution14:49
hazmatie. a new rel added post the invoker setup, or one removed14:49
jimbakerhazmat, there's testing of adding a relation. there should be one for removing a relation, as you mention14:49
jimbakerhazmat, i can readily add that14:50
hazmatjimbaker, can you point me to it?14:50
hazmatjimbaker, the add one that is.. i'm also wondering about membership consistent views on membership mutation14:50
hazmatlike if one relation has a unit, and another should have it, but its been removed during the execution.14:50
jimbakerhazmat, see test_get_relation_hook_context14:51
jimbakeri didn't address membership14:51
hazmatjimbaker, yeah.. its tricky, we don't want to fetch all units of all rels for each hook exec, but it does make a hook's perspective on membership possibly inconsistent when working across multiple relations14:54
jimbakerhazmat, yes14:54
hazmatjimbaker, so effectively we do already know the memberships14:58
hazmattheir in the respective schedulers14:58
niemeyerOops.. no  Go reviewers..15:03
jimbakerhazmat, the scheduler does pass in the membership to the implied relation hook context15:03
jimbakermy proposal continues the current practice of caching it on first read, in this case as would be done with relation-list15:04
niemeyermthaddon: Can you please check this out and include a LGTM in case it actually Looks Good To You? https://codereview.appspot.com/596706415:11
mthaddonniemeyer: can you update the RT with what you want me to do? - currently otp15:11
niemeyermthaddon: Cool, nevermind.. will just submit as I have to leave for lunch in a few minutes15:12
hazmatjimbaker, sure.. but we could have consistent membership if we could pass the other schedulers membership sets to their respective hook contexts.15:12
hazmatjimbaker, i'm fine with that being in another branch, but what do you think of the approach?15:13
jimbakerhazmat, there's still no guarantee that you have a consistent view of membership - only that you have the other schedulers cached membership15:20
hazmatjimbaker, the schedulers cached membership is point in time consistent with what the unit has been informed of to date15:21
hazmatits altered as various hooks are executed.. hmm.. although the execution representing the membership change may merely be queued, such that the membership change notification is pending but not actualized/known to the charm even though its represented in the cached set.15:23
jimbakerhazmat, right, scheduling is occurring differently than membership watches - there are effectively 2 queues here15:26
hazmatjimbaker, um.. not true per se.. scheduling is the target of the membership & change watchers, but that's separate from execution. the cache is consistent wrt to execution of hooks for a single context, but its altered prior to the execution.. that's a trivial to fix.. it can do the cache alteration after successful exec and pass in the modified set to the execution, and then it could be used as a membership set for other contexts.15:28
hazmatjimbaker, but yes there are two queues.. the scheduler queue, and the hook executor queue15:28
niemeyermthaddon: Change submitted, config file support is ready, RT updated15:29
niemeyerand lunch time15:29
niemeyerbiab15:29
mthaddonniemeyer: cool, thx15:29
niemeyermthaddon: np, I'm back too16:23
niemeyermthaddon: Let me know how it goes, please16:23
hazmatjimbaker, is there a reason why the relation-id-option has a merge target of the relation-hook-context?16:52
hazmatbcsaller, this reitveld diff looks borked.. like trunk hadn't been merged yet to the parent/prerequisite https://codereview.appspot.com/5845061/16:54
jimbakerhazmat, that's mysterious to me... i simply used lbox propose -req to specify that it required lp:~jimbaker/juju/relation-ids-command as prereq16:54
jimbakerdoes lbox work with multiple levels of prereqs?16:54
jimbakerthat was my assumption, since we use that in lp16:55
bcsallerhazmat: yeah, it was merged before, but I'll look at it again, I noticed this this morning as well16:55
hazmatjimbaker, it justs a pre-req, and does the diff against that, but it does have a -target option that result in changing the merge target from its default, but the default may indeed be the branch's pull location16:56
hazmatnot sure16:56
jimbakerbcsaller, have you been able to use lbox for stacked branches?16:57
bcsallerjimbaker: my branches got so tangled up its hard to know where the issue is :-/16:58
bcsallerjimbaker: so maybe... :)16:58
jimbakerbcsaller, understood16:59
hazmatjimbaker, are you using bzr-pipeline?17:02
jimbakerhazmat, i do not17:02
hazmatbcsaller, the status changes look very nice17:02
bcsallerhazmat: cool17:02
hazmatbcsaller, can you re propose it via lbox to clean up the diff?17:03
bcsallerI don't know if that will clean it up, but I can try17:03
hazmatbcsaller, assuming the parent/pre-requisite has parent merged, it should17:03
hazmater.. has trunk merged17:04
hazmatoff to an appt, bbiab18:13
jimbakerbiab, my son fractured his wrist and needs a cast put on20:02
niemeyerTime to reboot..21:27
niemeyerHmm.. things looking better in Unity now22:11
niemeyerI wonder if the network issues have been fixed too22:11

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!