/srv/irclogs.ubuntu.com/2017/02/13/#juju.txt

AnkammaraoHello Juju World!!!!!07:03
Ankammaraoi want to know , is there any command or process to remove or hide the old charm verions in the charm store07:03
kjackalGood morning Juju world!08:08
kjackalAnkammarao: I believe you can move the charm you have to an "non searchable" channel08:08
Ankammaraokjackal, how do i move , is there any command or process08:39
=== frankban|afk is now known as frankban
kjackalAnkammarao: you could try charm grant to remove access of a revision on a channel09:10
Ankammaraokjackal, i have tried with the grant command by giving access to only one user , but other users alos able to see now09:17
kjackalAnkammarao: Reading the help of charm grant it seems you need to use the --set flag to overwrite any already existing acls09:21
kjackallike so: charm grant ~johndoe/wordpress --acl write --set fred,bob09:21
kjackalAnkammarao: There is also the "charm revoke" you can try09:22
nobutohi, can somebody review this pull request of apache-php layer? https://github.com/juju-solutions/layer-apache-php/pull/409:23
kjackalAnkammarao: you can do a "charm revoke --help"09:23
anita_Hi I want to revoke the grant for one charm for specific versions only09:52
anita_when I tried with charm revoke, revoking happening charm wise09:53
anita_not version wise09:53
Ankammaraokjackal , i have tried with the revoke command , its removing or hiding complete charm , but here we want to hide only the older verions not entire charm10:01
kjackalAnkammarao: so revoke will not work if you pass in the revision like this: "charm revoke ~johndoe/wordpress-4"  ?10:09
kjackalAnkammarao: Then i think charm grant with the --set on the revision is the only option you have10:11
kjackalAnkammarao: like so: charm grant ~johndoe/wordpress-4 --channel edge --acl write --set fred,bob10:11
Ankammaraokjackal, so only fred,bob can see the charm revisions or version10:15
kjackalAnkammarao: yeap, based on the documentation10:16
=== plars-away is now known as plars
=== perrito667 is now known as perrito666
=== zerick_ is now known as zerick
Hetfieldhi, i'm using LXD local cloud provider for juju15:25
Hetfieldi would like to enable controller HA by enable-ha command on 3 different machines, but always LXD.15:25
Hetfieldi don't know how to tell juju to use a distributed provider like that15:26
Hetfieldwhile using MAAS, instead, it's easy and working15:26
rick_hHetfield: that's not currently supported.15:29
rick_hHetfield: for the moment, juju only works on one machine with lxd at a time. There's work that's been talked about for making lxd more like a cloud and juju would be updated to do more what you're looking for15:29
Hetfieldrick_h: oh, good, this means i'm totally dumb :)15:29
rick_hHetfield: no, means you're smart for wanting it and we're working on it :)(15:30
Hetfieldrick_h: actually it would just mean to add more endpoints or model it as regions15:30
Hetfieldrick_h: any ETA?15:30
rick_hHetfield: now you're thinking. More like availability zones tbh15:31
rick_hHetfield: I think the lxd bits are looking to be in 16.04 and then juju would update in the next cycle15:31
Hetfieldrick_h: ok. but nowaways how to get HA for production usage (on premise)? only way is MAAS if i understand good15:32
rick_hHetfield: well maas, any public cloud, openstack, etc all support HA on different hardware like that15:33
rick_hHetfield: lxd is meant to be a local/one machine so HA on that isn't a thing. It's more the exception to the rule tbh15:33
Hetfieldrick_h: agree15:34
rick_hHetfield: maybe look at HA at the application level across different machines running lxd? e.g. if one machine goes boom the others can keep applications going and be brought back up?15:34
Hetfieldrick_h: i meant infra HA. i mean...if machine with juju-controller ( i will deploy openstack-charms) dies, what happens?15:34
rick_hHetfield: so you're running openstack on a single machine?15:35
Hetfieldi didn't want to add pacemaker or such cluster tools15:35
Hetfieldso, on several, let's say 1015:35
rick_hHetfield: sorry, I think I got confused. So you're going to run openstack then yes, we'd suggest you use maas to get that going15:35
rick_hHetfield: if you want to put the controllers in containers you need to manually setup VMs on the machines you want and add them to maas15:36
rick_hHetfield: then target those in bootstrap/enbale-ha with the --to/constraints arguments15:36
Hetfieldoh, yes, it was my solution15:36
rick_hHetfield: there's not really a suggested production way to do openstack with the lxd provider, more the maas provider and our default tools stick things in lxd containers and spread them across machines for resiliency15:36
Hetfieldbut as i'm going to use lxd for openstack components (apart nova...) i wanted to add lxd juju controllers too15:36
rick_hHetfield: right, there's the chicken and egg problem there. To get around it you have to have the containers setup and in MAAS that looks like machines15:37
Hetfieldyes right15:38
Hetfieldit's very interesting15:38
Hetfieldrick_h: a sort of: how to compile gcc :)15:38
rick_h:)15:39
kklimondais there documentation on creating local mirror of juju agent (and probably other tools) needed for bootstrap? I have a terrible internet connection where my juju is installed at.15:54
=== balloons26 is now known as balloons
kklimondalooks like I have sstream-mirror https://streams.canonical.com/juju/tools [...]16:03
kklimondabut I'd rather avoid downloading 5.5GB with all the releases16:04
ZiclazyPower: hi, if you have any things that I can beginning to test tomorrow morning without waiting for you (as we don't have so much time where we can talk, both of us, in a day with the local clock time) :)17:13
lazyPowerZic - i just got a good build of the charm deployed in lxd (1 time)17:13
Zicdat sync \o/17:13
anita_Hi, How can I revoke one specific version of a charm?17:14
lazyPowerZic - by end of day i'll have pinged you with a pr  and a personally published charm revision you can use for testing17:14
anita_i tried revoke, but it is revoking charm wise17:14
lazyPowerwe might push this in the edge channel i need to sync with the team17:14
lazyPowerZic - however, the end game is we are on the right path to getting this fixed for your use case by end of week17:14
Zicthey don't call me today (I'm out of office), so I expect the cluster worked fine with the singlemaster without me :)17:14
lazyPowerZic  fantastic :)17:14
lazyPowerthanks for pinging back and keeping up on this17:14
lazyPoweronce i've got this in flight in kubes/kubes, i'll need your feedback on the PR and i'll need to rotate to another issue, however this should get yo moving for more testing/feedback of the HA master. There's likely to still be some gremlins in there as this is first-pass stuff17:15
lazyPowerwell second-pass actually17:15
lazyPowerbut i digress17:15
anita_Hi, How can I revoke one specific version of a charm? i tried revoke, but it is revoking charm wise, all versions are revoking17:20
lazyPoweranita_ - you can change the published "tip" of that charm stream to a prior revision or the next revision in sequence17:22
lazyPoweranita_ charm release cs:~team/entity-version  --channel=stable   should move the stable revision where you're looking to point it.  same with other channels, just sub in the channel name17:22
anita_lazyPower_:I didnt get17:22
lazyPoweranita_ - however i think the point is you dont revoke, its a moving "alias" pointing at the revision you want to represent the latest known good revision for that channel17:23
lazyPoweranita_ - as teh charm itself is all tracked linearly in the "Unpublished" stream. you're just pointing to different combinations of the charm version and potentially any resources, when you do the charm release. there' s a diagram for this in the docs under the charm store page in the dev guide17:24
anita_like I want to grant 6th version where as i want to revoke 4 and 517:24
lazyPoweranita_ https://jujucharms.com/docs/stable/authors-charm-store#entities-explained17:25
anita_lazyPower_:Ok let me check17:25
ZiclazyPower: feel free to PM me in IRC if instructions are too long to put it here :) I will begin test tomorrow at 09:00 UTC, I will have time to do some tests before you wake up and make a recap of what I observed ^^17:34
lazyPowerZic - sounds good17:34
lazyPowerZic - i'll put it somewhere thats simple like juju deploy cs:~lazypower/canonical-kubernetes --channel=edge  or somethign similar. the idea is to get it in your hands as fast as possible17:34
lazyPowerso i'll eat that pretext work, and you focus on the usability/feedback cycle17:34
lazyPowerwhen you're ready we'll fold you in on the buiding from source routine, its a little complex for first time juju developers... a lot of custom tooling in here17:35
Zicmy time is divided at work between CDK/Kube and learning Go this days... so I'm available for intensive tests if you need :)17:35
lazyPowerZic but if i can get you knowledgeable in our stack, it would be great to add you as a reviewer of the CDK code as it comes up for landing in kubernetes/kubernetes17:35
lazyPowernot that i have ulterior motives here ;)17:36
ZicI'm interested in it yeah, as I'm planning to write some juju charms in a near future17:36
Zic(and all knowledge about Juju will be helpful, as one of my teamworker is planning to use Juju for an OpenStack cluster)17:36
anita_lazyPower_: I want to revoke the read/write permission completely from those versions.17:38
lazyPoweranita_ - there is no revoke, you can change the head pointer17:38
anita_oh ok17:38
lazyPowerthe grant/revoke will sweepingly add permissions to an entire channel17:38
lazyPowerso thats why it appears like you've revoked the entire charm. Thats intended for doing early access isolation to a group of users for testing, eg: edge can only be deployed by dept. a and dept. b17:39
lazyPowerdept c - f will have to wait until it lands in candidate, as its too high risk for those models17:39
anita_ok17:39
lazyPowerthe whole point is risk assessment and subscription, the reason for no revoke is if someone deploys that revision and you revoke it, you've abandoned htem in the release chain17:39
ZicI don't know if Juju's code is a great place to show "Go in action" for a very-beginner of Golang :> but one of the attribute of Go seems to be "you will be able to read other's code in a week"17:40
Zican attribute that, as a C developer, I was never completely able to, after some years of C, depending of the software-stack17:41
Zicwe'll see :D17:41
lazyPowerZic - well I was also told ruby will eat teh world of software and we see how far that went17:42
Zichehe, I'm still confronted to Ruby as we heavily use Puppet here17:44
Zicand for some hacks that cannot be in Puppet's language, we directly neeed Ruby's code :(17:44
lazyPowerZic - i'm familiar with the syndrome17:45
lazyPowerthats one of the reasons why we went with reactive and pure python. DSL's are great when you *want* rails, but there's often times where you need to write up a funky little method to do something highly specific and it fits together better like that.17:45
ZiclazyPower: I was told that JS will be everywhere, as Web Developers "contaminate" (with all my respect to web developers) the system-programming with NodeJS17:45
lazyPowerartisinal coding vs cargo-culting.17:45
Zicand now, we see Rust, Go and other compiled language, with a strong approach to be system-programming17:46
Zicso I'm happy :)17:46
lazyPowerZic - yep, predictions are just that, marketing tools to guide you somewhere, despite the social dynamic and "social climate" being very different indicators of that statement.17:46
Zic*execept* when I saw that Unity 8, and even GNOME project emphasis JS as the "de facto primary language" :'(17:46
lazyPowerwell you never know ;)17:46
Zicwith Unity 8 using pure-QML, and GNOME with GJS17:46
lazyPowerjs does make for an awesome wiring language to put together UI elements and give them behavior17:47
lazyPoweri'm kind of happy to see ECMASCRIPT getting some love17:47
lazyPowerbut thats a very different beast than the devops we're working on17:47
Zicin my company, most of our customer stills use PHP for backend, some of them used NodeJS17:48
Zicsome of *our* backend are in Go at contrary :)17:48
lazyPowerwell with the incoming dockerpocalypse we'll see a lot more using the whole menagerie of backend tech because containers and their nature of disposable infra17:49
lazyPowerfor quick POC style testing, and the longer-term applications that make the cut17:49
Zicwhen I talk about LXD in our weekly meeting, the immediate question was "And can we run Docker in it ? And in what way it's different from Docker in Docker?"17:49
lazyPoweri woudln't be surprised if you find a stack of python, ruby, nodejs, go, php, and probably even some erlang if you're persnickety17:49
Zic*sigh*17:49
lazyPowerZic - we have some material on that which may help17:50
lazyPowerthere's an inforgraphic for lxd vs docker17:50
lazyPowerwhich calls out the strengths of each given context17:50
lazyPowerZic - something like this is a good way to start teh conversation https://insights.ubuntu.com/2015/09/23/infographic-lxd-machine-containers-from-ubuntu/17:50
Zicyeah, but I know it will finish in "LXD is replacing our VMs, right, but... *Where can I run my Docker?*"17:51
lazyPowerZic - the answer is "just about anywhere/everywhere"17:51
lazyPowerwe've gone through great lengths to ensure docker works in all those substrates17:51
Zicthe one thing is at least, I will try to replace my local use of KVM/Qemu with LXD containers for my future test17:52
lazyPowerwe want docker on ubuntu to just work. I think the only place left as a hold out is bash on windows, and who knows where that API is, i'm not working on the project so its hard to say.... but if i were to guess those system calls will eventually be translated and you'll see something fun in there.... but dont hold me to that, I have no idea if thats on the roadmap, its pure speculation.17:52
lazyPowerZic  - man ;) I'm using lxd right now to test this HA patch17:52
Zic:D17:52
lazyPowerthe brilliant part of LXD is the capacity to model and test HA deployments locally17:52
lazyPoweryou can *even* simulate network partitions17:53
lazyPowerwith a simple profile edit on the lxd container17:53
lazyPowerwant to see what etcd really does when you pull the plug on 2 units? no problem17:53
lazyPowersnip snp17:53
lazyPoweroh look 2 units died and the one unit is complaining it cant find quorem17:53
lazyPowerwelp,  i guess 3 nodes is disaster, better scale to 517:53
lazyPower(how i validated the etcd documentation assertions)17:54
Zicanyway, what push me to do some learning of Go was because Canonical's tools seems to use Go as primary language now (was Python few years back); before I was like "meh, another language... at least it's compiled, fine" :>17:54
lazyPowerlol i nkow you're excited about this17:54
lazyPoweryou were all jazzed last week17:54
lazyPoweri'm sorry i dont share your enthusiasm about go :) I'm pretty lazer focused on our k8s roundup these days17:54
Zichehe :D17:54
ZicI'm not a go-enthusiast for now, I'm still learning it as "secondary-class language", like I did with Ruby (was forced to, with Puppet)17:55
Zicmaybe it will become one of my main tooling language at future17:55
Zicmaybe. :)17:55
lazyPowerZic - the world can only hope for so much awesomeness my friend17:56
Zicfor now, my main language are Bash, Python & C17:56
lazyPowergo forth and conquer17:56
Zicand I didn't talk about Rust :D as I'm a pro-Firefox, I'm stalking Servo evolution and Rust language too :)17:58
ZicRust is not as easy to adopt as Go anyway, but seems to be a great language also17:58
lazyPoweri know the prime author behind it. I used to attend meetups with steve klabnick18:00
lazyPowerhe's a pretty smart dude, so i have no doubt rust is the bees knees18:00
Zicit's one of the reason I love C, it's nearly 45 years now iirc, it resists to the "hype and the dance of programming language"18:01
ZicI love to saw one of my sandbox program of 1999 still running and that I'm able to edit it easily18:02
Zicit's not the same when I found an old Java program of myself :p18:02
Zic(yeah, I did Java... at school... *burp*)18:02
stormmorehowdy juju world!18:30
=== frankban is now known as frankban|afk
lazyPowero/ stormmore19:15
lazyPowerZic making the PR now, will ping you with instructions before i move to the next objective19:15
derekcatHey everyone, does anyone know where Juju stores ssh known_hosts? It keeps telling me to:  ssh-keygen -f "/tmp/ssh_known_hosts736182584" -R <IP address>  ...Which obviously doesn't work when the /tmp version disappears...19:25
lazyPowerZic - if you want to tag and follow along - https://github.com/kubernetes/kubernetes/pull/4135119:34
CarlFKhow do I clean up machine 1?   juju status: http://paste.ubuntu.com/23990363/19:36
lazyPowerCarlFK juju remove-machine 1 --force19:36
CarlFKlazyPower: thanks. one down.  now to figure out why  30 min later cnt5 is still "installing charm software"19:38
CarlFKnever mind.  need for it has evaporated.19:40
lazyPowerZic - simple instruction is to deploy teh bundle via conjure-up as normal, when you're at the waiting/allocating screen: juju upgrade-charm kubernetes-master --switch cs:~lazypower/kubernetes-master --channel=edge19:41
lazyPowerZic - that should replace your masters with the patched version for HA and you should be able to start testing from there. it's in the pipeline and ready for testing from my store copy of that charm. The remainder of the bundle was untouched by this change.19:42
marcoceppiderekcat: I think that's on the controller? rick_h ^?20:08
derekcatmarcoceppi: Do you know where it might be on the controller?  I've tried find / -name known_hosts but nothing is found..20:10
derekcat*as root20:10
bdxderekcat: I don't think there should be a known_hosts file by default, should there be?20:19
bdxjuju just adds ssh keys to .ssh/authorized_keys for the ubuntu user to allow access20:19
derekcatbdx: It's got one somewhere..  From the box that I administer Juju from, I can ssh ubuntu@<IP of machine added to Juju>, but when I try to juju ssh postgresql/14 it comes back with a host identification error: @    WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!     @20:23
marcoceppiderekcat: is there one in ~/.local/share/juju/20:30
marcoceppi?20:30
derekcatmarcoceppi: jujuadminbox:~/.local/share/juju/ssh only has juju_id_rsa  juju_id_rsa.pub20:33
marcoceppiderekcat: I've not been able to find this either. It might be worth asking on juju@lists.ubuntu.com20:33
derekcatmarcoceppi: Ahh dang..  I'll give that a shot.  Thank you!20:35
GMR-OBAnyone in here who knows how to fix a Percona Cluster (JUJU CHARM)  when all nodes are in blocked status ?20:38
lazyPowerGMR-OB - existing deployment or fresh deploy?20:41
GMR-OBexisting Deployment happened after a RAM crash on a single Server20:41
GMR-OBBox is back up but now Percona Cluster is  saying status blocked on all 3 nodes : juju giu shows all ok20:42
lazyPowerGMR-OB  is this part of an openstack deployment?20:42
=== GMR-OB is now known as teranet
teranetyes it is sorry had to fix my nicname too LOL20:43
lazyPowerteranet np :) I was going to suggest poking in #openstack-charms, i would suspect that whatever is blocking you has come up in CI before20:43
teranetok will do thx20:43
lazyPowerand its likely someone there might have some advice, otherwise i'm happy to take stabs at helping you resolve with generic troubleshooting advice20:43
ZiclazyPower: ack, I will try this tomorrow :)20:54
lazyPowerZic - final note is its cs:~lazypower/kubernetes-master-1120:54
lazyPowerif you see it grab a diff revision, somethings wrong and you should specify revision 11 explicitly to the command. but the channels should "just work"20:55
=== scuttle` is now known as scuttlemonkey
stormmoredamn it getting a bad gateway error :-/23:50
lazyPowerstormmore - on cdk?23:54
stormmoreyeah but it is most likely my fault :-/23:55
lazyPowerstormmore - ive had some reports of 502 bad gateways on deployments recently and i haven't been able to reproduce23:55
lazyPowerif you can reproduce it reliably please file a bug, it might be racey23:55
lazyPowerwhich would explain why i'm not seeing it and others are23:55
stokachustormmore, you deploying on aws?23:57
stormmoreI just did a basic ingress controller but I suspect it is cause I got my container to redirect http to https23:57
stormmorestokachu yeah I am23:57
stokachustormmore, what version of juju?23:57
stormmorejuju 2.0.323:58
stokachuhmm ok23:59

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