=== bogdanteleaga_ is now known as bogdanteleaga === TheRealMue is now known as TheMue === StoneTable is now known as aisrael === ryotagami_ is now known as ryotagami === mup_ is now known as mup === Lukewh_ is now known as Lukewh === tris- is now known as tris === skr is now known as Guest61781 [07:03] Hi all, I added my openstack cloud to juju clouds and credentials also. while doing juju bootstrap to that cloud i am getting following error. [07:04] Error log pasted here http://paste.openstack.org/show/595548/ [07:04] please someone help [07:27] surf: what's the version of openstack? [07:27] axw: liberty [07:32] surf: can you share your clouds.yaml? [07:33] axw: yeah 1 sec [07:33] here is my clouds.yaml file http://paste.openstack.org/show/595553/ [07:33] thanks [07:34] surf: I'm a little suspicious that we're not doing v3 properly. can you try changing the endpoint from v3 to v2, see if that makes a difference? [07:35] axw: ok [07:36] awx: still no difference [07:38] surf: did you set anything for "domain-name" in the credentials? [07:38] yeah it is "default" [07:39] surf: I have to admit to being completely ignorant of v3 auth, I've never used it. can you please try commenting that out of your credentials.yaml to force v2 auth? [07:40] if that doesn't work, might need to wait for someone with better openstack debugging skills === frankban|afk is now known as frankban [08:23] axw: still i am getting autentication problem [08:31] Good morning Juju world [08:33] surf: is the debug output any different at all when using the v2 endpoint? [08:41] axw: here is the debug log with v2 http://paste.openstack.org/show/595565/ [08:47] surf: sorry, I am out of ideas [08:48] axw: ok thanks for your i nfo [09:47] Hi all, I am doing bootstrap juju to the openstack cloud [09:48] is it necessary the both juju and openstack cloud on same node [09:49] or can i bootstrap juju on to the openstack cloud from different machine, where juju is installed === axino` is now known as axino [11:53] hi all, I followed this doc https://jujucharms.com/docs/devel/howto-privatecloud [11:53] to bootstrap juju on openstack cloud [11:54] i am getting the error authentication failed. [11:54] here is the debug log http://paste.openstack.org/show/595606/ [12:11] surf: are you deploying whole openstack or just creating controller to existing openstack tenant? [13:00] Hey cory_fu do you happen to be around? [13:05] hello there, is it possible to have some charms wait during a bundle deployment for after a specific other charm is finished? [13:16] BlackDex: "waiting" logic like that would need to be in the charm itself [13:17] for example, you could wait to start the service that the charm provides until after it had successfully joined a relation with another charm [13:18] e.g. a web app charm that's waiting for a database [13:27] well for instance if i deploy a system with a lot of LXD containers, or even deploy mysql and wordpress onto a physical system without containers they both will install packages [13:27] since just one instance of apt-get can run, one will wait [13:28] if the juju deploy (using a bundle) is smart enough to wait to add one untill the first is finsihed, that would be better i think [13:32] BlackDex: is this theoretical or are you hitting a problem in a deployment? [13:33] hitting a problem [13:33] i just received an lxd container within a 10.x.x.x network, will the network is 192.168.x.x [13:34] so i needed to re-deploy [13:34] also, i'm currently trying to deploy an HA OpenStack [13:34] it seems like to much is running at the same time [13:34] some hooks fail [13:35] and if i run those hooks via debug-hooks there seems to be nothing wrong, or causeing errors [13:36] BlackDex: anything useful in `juju debug-log` output? [13:37] i have to check that a bit better [13:37] but not really [13:38] i this case i see that mysql isn't able to connect [13:38] but it is strange if i do it manually it works [13:39] BlackDex: which bundle are you deploying? [13:39] percona-cluster [13:45] BlackDex: so not a bundle, just the percona-cluster charm? [13:54] no a whole bundle [13:54] the basic openstack-bundle [13:55] and i changed it to be HA [13:55] the base works fine [13:55] but after ha it stops deploying even the basic charms them selfs [13:55] it looks like some limits or something [13:55] or race-conditions [13:56] ill go and try the default bundle now again [13:56] and see what that does [14:18] it looks like juju 1.25 is at least sometimes setting up /etc/apt/apt.conf.d/99no-recommends to set apt to not install recommended packages [14:18] is there a setting to enable/disable that? why is that happening? [14:31] that seems like really wrong behavior to me [14:31] yeah [14:31] it seems crazy [14:32] we're seeing breakage because base layer just installs python3-pip, which ends up bringing python3.5-dev in via recommended packages if apt is allowed to [14:32] but with no recommends set it doesn't, and then pip install of pyaml fails because it wants to build a cpython module [14:32] I'm not a policy expert but that seems like a bug to me [14:35] jhobbs: where are you seeing this no-recommends set? could it be a different colocated charm? Is something integrating with a charm before its deployed? I've never seen this, but I always wanted it :) [14:36] the file says it was created by juju [14:36] but it may be a lie [14:36] i just checked the juju source and don't see it listed [14:37] jhobbs: what charm? [14:37] ah ok [14:37] it's IS's basenode that is doing this [14:37] jhobbs: is there an exec.d dir in the charm root dir or in the charm hooks dir? [14:37] sorry for the false juju alarm [14:37] i will take it up with them [14:45] whew! [14:59] question concerning juju beta hosted controller [15:00] the "Access Last connection" doesn't seem to populate corrrectly see -> Access Last connection [15:00] oops [15:00] http://paste.ubuntu.com/23828205/ [15:01] juju-gui shows https://cloud.githubusercontent.com/assets/5590546/22111375/f9e80a5c-de13-11e6-92d3-1672174dd3b4.png [15:02] I'm wondering if this has to do with the permission level I have as a user of the controller ... [15:03] tvansteenburgh: With the old juju i used the juju-deployer and that waited with adding units after the first unit whas running, it also waited with adding the relations untill the end.. Since juju 2.0 it deployes better like juju-deployer, but it laks the waiting part. I think that is a plus [15:05] the waiting part i mean, that is better, it doesn't hogg the system all at once [15:06] that may be true for a local lxd deploy, but it's irrelevant for a typical cloud deploy [15:09] why? because all the HA services are within an LXD [15:09] all services are within an lxd container on several nodes [15:10] thats the same from my point of view [15:12] lxd containers are totally isolated from each other, so there shouldn't be any conflict (like using apt at the same time on multiple containers). and if the node hosting the lxds can't handle running them all at once (i.e. the system is overloaded), then the node is under-provisioned to begin with [15:15] well, there are a lot of tweaks needed by default to run lxd's like open-files, inotify, max-pids etc... [15:16] those arn't set by default high enough sometimes [15:16] it happens most on bare-metal systems [15:16] so multiple charms bare-metal on the same system. [15:17] i think i have found a way [15:17] i deploy it with hacluster charms, but i don't add all units yet. I created a small script to add those afterwards [17:05] kwmonroe: petevg: this is a fix for the path. Please, review, merge re-push to store. Apologies: https://github.com/juju-solutions/layer-cwr/pull/41/files [17:10] lol.. i love it. python nested in bash functions that are called by trapping the exit from our bash defined in a jenkins job xml. more of this please! [17:10] merged [17:20] kjackal: petevg: cs:~juju-solutions/cwr-30 is released (includes latest matrix, jenkins, and xml branches) [17:21] thank you! === frankban is now known as frankban|afk [20:15] Anyone have experience with juju failing to remove a unit (log shows "cleanup failed for removed Unit("unitname"): transaction aborted") [20:17] kwmonroe: agreed lets keep the python scripts under scripts. Thank you for releasing cwr [20:18] aisrael - i've only seen that deadlock when a remote relation has failed or a sub relation has failed [20:18] sounds like perhaps the agent cleanup is what failed? [20:19] lazyPower: Yeah, it seems like it. status shows the agent is stuck "executing". [20:19] aisrael if you juju ssh into that unit and recycle the jujud process, does it unstick it? [20:23] lazyPower: No, but it does show me a few things possibly broken in its networking, so that could be a clue [20:23] weird [20:23] i'm at a loss for why its doing this :( sorry [20:25] no worries, man. [20:26] * aisrael breaks out strace [20:27] aisrael: sounds like you need more mongos [20:28] I think I need eleventy of them, right? [20:30] i read that as mangos. everyone needs more mangos [20:33] hah. I rebooted the unit and it cleared something up because the destroy finally finished [20:43] Huh. Cleanup failed with an unexpected error: unknown operation kind run-action [21:08] aisrael - did it have actions queue? [21:09] it might be that we've hit a not as well tested path with teh agents where an action is scheduled, yet cleanup attempt ran before the action was completed, and now its fighting for what should be happening [21:09] i'm guessing here, but it sounds reasonable [22:09] in the juju bundle yamls, when you build them, can you use 'tags' as constraints for machines? [22:09] rmcadams: yes [22:10] https://www.irccloud.com/pastebin/uRqlOAh1/ [22:11] so under constraints, I can put tag=compute [22:11] as an example? [22:11] rmcadams: yeah, heres an example: https://github.com/ubuntu-openstack/zopenstack/blob/master/bundles/zkvm/xenial-mitaka-2-machine-control-plane-next.yaml#L25 [22:11] tags= [22:11] nice [22:12] building my own derivative of the openstack bundle to build a big configuration of six compute, six storage and then some infrastructure nodes [22:14] cool :) [22:15] just trying to learn the bundles, seems important ;) [22:15] hah yeah, you dont want to be deploying that stuff manually [22:15] its like deploying kubernetes charms manually, only with more relations and components [22:16] rip [22:16] i concede, openstack is the final boss of charming [22:17] beisner - i pay homage to my fore-fathers publically ^ [22:21] * beisner ^5s that k8s work, good sir [22:22] <3 [22:37] won't somebody stand up for the elephants in the room? [22:38] big data is where it's at. [22:39] thats more of a balrog [22:39] soz [22:40] kwmonroe you do make a good point [22:40] kwmonroe but i would have expected you to be the champion and remind me of my roots :P [22:40] java is *always* the final boss [22:40] k8s and the 'stack are great, but when you want to actually do real work (like wordcount), hadoop is where it's at. [22:40] Did someone say Java? [22:41] lol [22:41] kwmonroe https://cdn.meme.am/cache/instances/folder481/53692481.jpg [22:42] that's awesome lazyPower [22:42] i cant take credit, its from an old monitorama talk https://jamessdixon.files.wordpress.com/2015/06/image.png [22:45] course, if you wnna do .. something like wordcount which isnt wordcount, but is very similar, then bigdata on openstack with lxd is where its at. right. right? [22:45] right [22:45] Java Java jav [22:45] a [22:45] yeah, because: bopensdata. [22:45] beisner: You know that pings me right? [22:45] Java [22:45] java [22:45] RIP bruzer's notifications and productivity [22:46] you guys! [22:46] java 2 se for workgroups [22:46] y u guys make fun of yava? [22:48] java - only slightly more stupid than rocket league [22:48] my 6 yo is addicted to rocket league [22:48] * lazyPower smirks [22:48] so is admcleod_, dont let him fool ya [22:48] raising a proper geek here [22:49] heck yeah beisner [22:49] i approve of this message [22:49] haha good one [22:49] yeah like ive even played it once, ever [22:49] this hour, you mean? [22:49] admcleod_ - my new obsession is ARK [22:49] cross between dayz, turok, and minecraft [22:49] guys,. plz. RL sux lol [22:50] oh no admcleod_ has reverted to internet gibberish [22:50] he must be having a RL withdrawl [22:50] ARK is so fun [22:50] justicefries ^^^^^^ OMG YOU'RE HERE [22:50] o/ [22:50] how goes it? [22:50] actually i havent played for a week or so [22:51] admcleod_ back in EU so our ping doesn't suck? [22:51] quite well! taking a slow WFH day [22:51] lazyPower: back weds [22:51] ack, we'll have to link up and pwn scrubs on weds [22:52] lazyPower: then i will backboard second hit airdribble you into oblivion [22:52] i mean yeah pwn scrubs [22:52] these new juju nouns are great. [22:52] juju airdribble-controller --oblivion [22:53] haha [22:53] admcleod_ we call that the mr miyagi [22:53] its an advanced move, but totally cool when its pulled off [22:54] yeah. i mean. obviously i dont play RL, but if i did, i would also have been watching tactics vids. [22:54] oh lazy question on the CDK, took a cursory glance and couldn't find it - what's the default cert expiry on what easyrsa is issuing? [22:55] 10 years, mbruzek fact check me? [22:55] yeah, he confirmed on the hangout [22:55] cool [22:55] java_max_int for sure [22:55] yes 10 years by default and you can make them longer [22:55] 10 yars [22:55] *years [23:00] http://www.2600online.com/assets/cart_yarsrevenge.jpg [23:00] classic [23:00] @beisner bring back the door games [23:01] time for some L.O.R.D. [23:01] or TradeWars 2002 [23:05] also beisner did you know you can play atari in your browser? http://www.virtualatari.org/soft.php?soft=Yars_Revenge [23:05] * beisner can't click that right now, please leave a message [23:05] * lazyPower leaves that right here in channel for safe keeping [23:06] *visions of all CI breaking for weeks* [23:25] beer proof keyboards