/srv/irclogs.ubuntu.com/2016/11/23/#juju.txt

beisnerkwmonroe, haha - clearly, layer basic attempted a config-get too early.  ;-)02:06
kwmonroei don't think so beisner.. i didn't show the whole log (which i have conveniently lost by now), but the charm had already gone through "installing charm software", after which i think should be fair game to config-get to your heart's content.  i think the issue was more with "error: connection is shut down", as in maybe config-get isn't safe to run if a unit's jujud can't contact the controller.02:25
beisnerkwmonroe, mostly joking.  but yeah, the charm can't be expected to solve tooling/agent errors.  i'd raise a core bug.02:26
kwmonroebeisner: i appreciate your jovialityness.  i also like your no-retry stance and hope to live in your world someday.  it's just so much easier to push those silly errors under the covers until even a retry isn't enough.  it makes debugging future-kevin's problem, and i don't care about that guy.02:28
beisnerkwmonroe, ha!  happy to provide levity.  well to be clear and differentiate:  i think failed hook retry is a good thing in user/production land.  it removes needless face-slaps from transient infra/interweb type issues.  it's in dev/test/gates where i think it is actually counterproductive.02:35
=== junaid1 is now known as junaidali
=== junaidali is now known as list
=== list is now known as junaidali
=== junaid is now known as junaidali
=== frankban|afk is now known as frankban
kjackalGood morning Juju world!09:12
=== zz_CyberJacob is now known as CyberJacob
=== CyberJacob is now known as zz_CyberJacob
AnkammaraoHi All11:09
AnkammaraoHi All11:50
MmikeHi, lads. Any news on when 1.25.8 is going to hit the archives?11:58
marcoceppiMmike: should be in the next hour, at most, are you pulling from the stable ppa or the Ubuntu archive?13:14
Mmikemarcoceppi, depending on the customer :) I'm asking for trusty so both are applicable13:15
marcoceppiMmike: it's available here https://launchpad.net/~juju/+archive/ubuntu/1.2513:18
Mmikeoh13:19
Mmikehaven't checked the ppa today13:19
Mmikethn marcoceppi13:19
Mmikethnx13:19
AnkammaraoHi15:13
cory_fupetevg: Added a reply to https://github.com/juju-solutions/matrix/issues/24 to clarify that I didn't mean that we should break pre-deployment.  I agree that that's a useful dev feature.15:15
petevgcory_fu: thx for the clarification. I'm +1 on the issue now.15:16
AnkammaraoHi All15:19
Ankammaraocan any on tell me to create terms15:20
Ankammaraoi am not aware of creating terms15:20
rick_hmattyw: you able to help Ankammarao?15:22
cory_fukjackal: Am I correct in remembering that you can build Bigtop charms with a layer option "bigtop_version: master" or similar to get a charm that deploys the cutting edge of Bigtop?15:25
kjackalthis is somewhat accurate. But we decided we will not be supporting this option/path15:26
kjackalSo if i were to guess i would say it will not work15:26
kjackalHowever, I saw someone asking for this in the bigtop list15:27
kjackalcory_fu: ^15:27
cory_fukjackal: Yeah, that was Amir.  I think it's safe to let him know that it's available as an unsupported dev option15:28
kjackalsounds resonable15:28
cory_fukjackal: Am I right in remembering the value being "master"?15:29
kjackalI do not remember right now, let m echek15:29
kjackalyes it is master16:36
narinderguptahi all i started seeing the error during deployment ssl.SSLError: [Errno 1] _ssl.c:510: error:1409442E:SSL routines:SSL3_READ_BYTES:tlsv1 alert protocol version16:39
narinderguptahttps://build.opnfv.org/ci/view/joid/job/joid-deploy-baremetal-daily-master/1171/console16:39
bdxcharmstore is down17:08
bdxto some extent17:08
bdxhttps://s11.postimg.org/d2frsznwz/Screen_Shot_2016_11_23_at_9_12_24_AM.png17:09
urulamaweb part is down atn, service is up and juju deploy still works17:09
urulamaproxy issues17:09
bdxurulama: ok, thanks17:09
cory_fupetevg: Hrm.  I didn't run in to any issues with a fresh run of rules.1.yaml on that branch of matrix with a fresh model.  :/  I guess matrix and unit logs will be necessary.17:12
petevgcory_fu: I just did a re-run, and didn't run into an issue, either. Except for realizing that I was on the wrong branch, which definitely isn't an issue w/ your code :-)17:12
cory_fuheh17:12
petevgI'm now running on the actual right branch. The end result should be a blank model, correct?17:13
petevgYay! That's what happened.17:14
petevgcory_fu: was going to give bcsaller a chance to comment before merging, but the branch looks good to me, and the tests run great, too.17:16
urulamabdx: should be back up17:31
petevgcory_fu: just left a comment on your PR. There is a test that needs updating (didn't see it until after I blew up .tox and rebuilt)17:41
bdxurulama: nice, thx17:41
mattywrick_h, sorry - I can now, but it seems like Ankammarao has gone?17:43
=== zz_CyberJacob is now known as CyberJacob
mattywrick_h, for future reference, we do have docs for terms: https://jujucharms.com/docs/2.0/developer-terms17:45
rick_hmattyw: sorry, was otp and trying to punt17:48
cory_fupetevg: blarg17:49
cory_fu:)17:50
mattywrick_h, no problem at all, I was afk and forgot to set my nick, I'll keep an eye out tomorrow17:50
cory_fupetevg: Updated17:50
cory_fuThanks for the catch17:50
petevgcory_fu: after running glitch, I also notice that we can get into a state where we can't reset. I think that this is more of an underlying juju issue than an issue with your code, though. Will save off logs so that we can file a bug.18:00
cory_fupetevg: I think I might have seen that.  Is it due to agents being in a "failed" state?  I added some additional checks to the health task to spot that18:01
petevgcory_fu: yes.18:01
petevgcory_fu: I see an error in the logs, but I suspect that it's a lie, and that the actual problem is that the juju agent is borked: http://paste.ubuntu.com/23523296/18:02
cory_fupetevg: So, that should now set the "health.status.unhealthy" state, but I'm not sure how to properly handle that.18:02
viswesnI am getting error "Failed to copy file" during bs_manager.booted_context() and then task goes for destroying the controller; - http://paste.ubuntu.com/23523247/18:03
viswesnhow to overcome this issue18:04
petevgcory_fu: I can get it unstuck by running "juju destroy-machine --force <int>", where <int> is the machine name.18:04
cory_fupetevg: This touches on bcsaller's comment on the PR and we eventually might want to move to managing entire models.18:05
petevgcory_fu: in any case, I merged your PR. We can open a separate ticket to figure out what we want to do when the usual way of resetting the juju model doesn't work.18:06
=== frankban is now known as frankban|afk
bdxhttps://s22.postimg.org/oc82s31g1/Screen_Shot_2016_11_23_at_10_59_09_AM.png18:56
bdxSOS ^18:56
bdxahh, finally logged in18:57
bdxlooks like my controller is super loaded18:57
bdxhttps://s21.postimg.org/9mg4bhwef/Screen_Shot_2016_11_23_at_11_00_08_AM.png18:57
bdxjesus18:57
bdxmarcoceppi, rick_h: I'm thinking I'm going to need a massive controller here18:58
bdxrick_h: per the rackspace xenial support thing ... I was able to negotiate a deal to give me xenial for stateless services, even though its not supported yet :-)19:29
marcoceppibdx: what version of Juju?19:46
bdxmarcoceppi: 2.0.1-xenial-amd6419:54
marcoceppibdx: do you have a lot of models/applications deployed?19:54
bdxmarcoceppi: http://paste.ubuntu.com/23523955/19:55
bdxI may of had +5-10 more machines deployer earlier19:56
marcoceppibdx: so a decent amount. This might be a bug in the controller, but you can do one of two things. The first is you can just deploy controllers on larger instance sizes. The second is to enable-ha on the controller which will also help load balance reqeusts19:56
bdxmarcoceppi: ahh, controller ha faciltates load balancing too? sweet!19:56
bdxahh, only of requests though19:57
marcoceppibdx: I'm like 90% sure it does, basically all the agents get the address of all the controllers, and will over time spread those calls around19:57
bdxok19:57
rick_hmarcoceppi: bdx yes, reads, but not writes19:57
marcoceppirick_h: \o/ thanks for the confirmation19:57
rick_hbdx: will be interesting to see is 2.0.2 helps at all19:58
rick_hbdx: there's a lot of discussion and work going in to watching that as folks push things with multi-models/etc.19:58
* marcoceppi eod's for some R&R, cheers o/19:59
rick_henjoy and have a good holiday marcoceppi20:00
bdxmarcoceppi: thanks as always for your insight, happy holidays!20:01
bdxrick_h: what work is going into 2.0.2 that you are thinking will help lighten the load?20:02
rick_hbdx: so 2.0.2 is cut but in vetting atm because there might be a regression on maas with it20:03
rick_hbdx: but it does some work to help with cpu load.20:03
rick_hbdx: our internal folks are chomping to get at it20:03
bdxnice, good to know20:03
rick_hbdx: but there's some other work going on around identifying storms of activity and working on how to prevent the controller from getting swamped20:03
rick_hbdx: and for the cycle as a whole, there's a big goal to reduce the cpu load, especially at idle, because there's more event handling going on than you'd think20:04
bdxrick_h: how is that being approached?20:05
rick_hbdx: so the idle work is that we've got long running controllers with different sized models and identifying what events is keeping the CPU going20:06
rick_hbdx: and so there's some work to move to pub/sub vs poll/etc20:06
bdxahh nice20:06
rick_hbdx: so that things are only handled when required vs old fashioned whole sale20:06
rick_hbdx: so there's some very low hanging fruit20:06
smgollerhey, so I'm trying to use juju 2 (currently 2.1 beta 1 I think, whatever's latest as of yesterday) with maas 2, and when i bootstrap, it keeps trying to talk to version 1 of MAAS' API, which has been deprecated. Any ideas on how to fix this?22:14
smgolleryeah, 2.1 beta 1 and maas is 2.1.1+bzr554422:15
thumpersmgoller: do you have the endpoint set correctly?22:36
thumpersmgoller: because that should just work22:37
thumperI think we even have CI tests to ensure it does22:37

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