| === thumper-afk is now known as thumper | ||
| === mwhudson is now known as zz_mwhudson | ||
| jseutter_ | does anyone know how to get juju working with hpcloud? I tried following the instructions on https://juju.ubuntu.com/docs/config-hpcloud.html but get an error saying "cannot create service urls" | 03:40 |
|---|---|---|
| lazyPower | jseutter_, what is your geo set to? | 03:45 |
| lazyPower | s/geo/region | 03:45 |
| === wedgwood is now known as Guest44693 | ||
| jseutter_ | lazyPower: Ah! I just changed that to az1-region-b.geo-1 and removed hpcloud.jenv, but still get the same error | 03:46 |
| lazyPower | Hmm, that was going to be my suggestion | 03:47 |
| jseutter_ | hm, looking closer, the error also mentions "access to these services is missing: object-store" | 03:47 |
| jseutter_ | I wonder if I need to enable swift in hpcloud somehow | 03:48 |
| lazyPower | https://bugs.launchpad.net/juju-core/+bug/1237011, jseutter_ have you read this bug report? | 03:48 |
| lazyPower | according to wallyworld, you dont need to enable those anymore. it shoudl work ootb | 03:48 |
| jseutter_ | lazyPower: hm. Is that because of a change in hpcloud, or something in juju? | 03:49 |
| lazyPower | I would assume juju | 03:50 |
| lazyPower | but I'm not certain of that | 03:50 |
| lazyPower | well wally does mention swift in a post, so you may still need swift for juju to function correctly, and that is the object store in openstack yes? | 03:53 |
| lazyPower | it sounds plausible | 03:53 |
| jseutter_ | Found the problem. | 03:54 |
| jseutter_ | doh, hit the next one | 03:54 |
| jseutter_ | the issue was that in hpcloud, you have to hit the "activate" button on object storage to get access to swift | 03:56 |
| jseutter_ | now I get "cannot start bootstrap instance: index file has no data for cloud {az1-region-b.geo-1 https://region-a.geo-1.identity.hpcloudsvc.com:35357/v2.0/} not found" | 03:59 |
| jseutter_ | hitting that page in my browser gives an xml response that says "Not Allowed". | 04:02 |
| lazyPower | jseutter_, just a suggestion, but check this AU post | 04:04 |
| lazyPower | http://askubuntu.com/a/166081/6807 | 04:04 |
| lazyPower | did you follow those steps to get started? | 04:04 |
| lazyPower | i'm not terribly familiar with HPCloud so this is a learning experience for the both of us | 04:04 |
| jseutter_ | lazyPower: heh. I started there first, but that askubuntu post seems older than the directions on juju.ubuntu.com that I referenced earlier | 04:05 |
| lazyPower | Ok, I'll make a note to get them updated. Thanks for that feedback | 04:06 |
| jseutter_ | like in that askubuntu post, hpcloud didn't have ubuntu images yet | 04:06 |
| jseutter_ | lazyPower: The juju.ubuntu.com page for hpcloud is fairly badly out of date as well. I can file a bug if it would help | 04:07 |
| jseutter_ | it looks like hpcloud recently upgraded the version of openstack horizon they are based off of | 04:07 |
| lazyPower | jseutter_, please do - https://bugs.launchpad.net/juju-core/docs | 04:10 |
| jseutter_ | lazyPower: yeah, it's my first time with hpcloud as well. I spent the day trying to get juju working with devstack, then gave up and tried hpcloud as a last resort | 04:10 |
| lazyPower | I've got juju working on Amazon and LXC providers, that's right around the end of my knowledge of the providers | 04:11 |
| lazyPower | I figure I'll be digging around in openstack very soon, but unfortunately I'm not much help to you right now | 04:11 |
| lazyPower | aside from moral support and extra google'ing | 04:11 |
| === thumper is now known as thumper-afk | ||
| jseutter_ | lazyPower: thanks, that helps. I'm working on a charm that talks to nova, so I need some sort of working openstack install | 04:26 |
| jseutter_ | I might try rackspace next | 04:26 |
| jseutter_ | heh, apparently rackspace is a non-starter | 04:27 |
| jseutter_ | lazyPower: https://bugs.launchpad.net/juju-core/+bug/1274379 filed | 04:39 |
| jseutter_ | hm, wonder who else offers openstack that will work with juju... | 04:39 |
| lazyPower | THe only listed providers are HP and Maas/openstack | 04:40 |
| lazyPower | the manual provisioner isn't completed | 04:40 |
| lazyPower | I'll reference this bug tomorrow in my standup and see if there is already WIP on the hpcloud docs being updated. | 04:41 |
| lazyPower | but for now i'm going to EOD, its well past quitting time | 04:41 |
| jseutter_ | ah, thanks. | 04:41 |
| jseutter_ | same here | 04:41 |
| lazyPower | if you need anything else, dont hesitate to ask jseutter_ | 04:41 |
| lazyPower | take care | 04:41 |
| wallyworld_ | jseutter_: lazyPower : sorry, i just noticed your posts. from memory, there is only images metadata for region a right now on hp cloud, so for region b you need to generate and use your own metadata | 04:54 |
| lazyPower | wallyworld_, for clarity can you point me to information regarding that? I ran across a prior transcript from IRC without instructions. | 04:55 |
| jseutter_ | wallyworld_: ah, I'll try region a then | 04:55 |
| wallyworld_ | i just manually inspected the json metadata file to confirm it only had region a in it. | 04:56 |
| wallyworld_ | i'm not sure of the current state of the doco | 04:56 |
| wallyworld_ | i can ask to see if the folks doing the roll outs can include metadata for region b | 04:56 |
| wallyworld_ | there's a getting started guide somewhere, i'll see if i can find a link | 04:57 |
| wallyworld_ | the only doc i know of is https://juju.ubuntu.com/docs/config-hpcloud.html | 04:58 |
| wallyworld_ | i haven't read it so am not sure if it's up to date | 04:58 |
| wallyworld_ | it's a little dated | 04:58 |
| wallyworld_ | you no longer need public-bucket-url or control-bucket if running juju 1.16.5 or greater i think | 04:59 |
| lazyPower | wallyworld_, ah yeah i've read through that. Sorry i thought there would be more information specific to generating the metadata | 05:00 |
| wallyworld_ | ah, i'm not sure if or where that's documented yet. what version of juju are you running | 05:01 |
| lazyPower | 1.17.1 | 05:01 |
| arosales | jseutter_, are you running 1.17.x? | 05:02 |
| wallyworld_ | here's the current draft doco, there's a bug to turn it into user doc http://pastebin.ubuntu.com/6841941/ | 05:02 |
| jseutter_ | wallyworld_: so I tried switching regions and the error is still the same. "ERROR cannot start bootstrap instance: index file has no data for cloud {az1-region-a.geo-1 https://region-a.geo-1.identity.hpcloudsvc.com:35357/v2.0/} not found" | 05:02 |
| arosales | lazyPower, if you are running 1.17.x in hp cloud you need to add tools-url: https://region-a.geo-1.objects.hpcloudsvc.com/v1/60502529753910/juju-dist/tools | 05:02 |
| arosales | 05:02 | |
| arosales | to your hp stanza | 05:02 |
| jseutter_ | arosales: 1.16.4-saucy-amd64 | 05:02 |
| arosales | jseutter_, ah 1.16.4 shouldn't need to the tools url | 05:03 |
| arosales | but lazyPower if you are running 1.17 on hpcloud you will need the tools url | 05:03 |
| wallyworld_ | lazyPower: also, 1.17.1 is supposed to automtically upload image metadata to the cloud storage if you generate it so the bit about uploading using swift can be ignored | 05:03 |
| wallyworld_ | arosales: we *really* need to get metadata/tools up on streams.canonical.com | 05:04 |
| arosales | jseutter_, and your getting a 404 in region a? | 05:04 |
| wallyworld_ | arosales: this was supposed to happen before 1.17 was released | 05:04 |
| arosales | jseutter_, remember if you switch to region a to activate both compute and storage for region a | 05:04 |
| jseutter_ | arosales: checking.. | 05:05 |
| arosales | wallyworld_, simple streams has been fun this week | 05:05 |
| wallyworld_ | yeah, especially for maas :-( | 05:05 |
| wallyworld_ | we gave roger a workaround though | 05:05 |
| wallyworld_ | if streams.c.com was up, that wouldn't have been an issue for maas either | 05:05 |
| arosales | wallyworld_, at least there is that | 05:05 |
| wallyworld_ | arosales: root cause there was maas can't return a url to a file unless it has already been uploaded | 05:06 |
| wallyworld_ | other providers don't have that issue | 05:06 |
| wallyworld_ | and we missed it cause we don't yet have CI testing for maas | 05:06 |
| jseutter_ | arosales: compute and object are activated for region a. When I hit the url from the error in my browser, I get an xml message that says "Not Allowed". Not sure if that's a valid test or not. | 05:06 |
| wallyworld_ | that xml message is cause tools/image metadata cannot be found - it is a poor message but is fixed in 1.17.1 | 05:07 |
| wallyworld_ | it would be goo to run bootstrap with --debug --show-log to see the output so we can tell where it's trying to find tools | 05:08 |
| arosales | wallyworld_, we need better ci with maas for sure | 05:09 |
| wallyworld_ | arosales: william and others are now going to push hard to get that happening | 05:09 |
| arosales | +1 | 05:09 |
| arosales | jseutter_, and you pastebin juju --debug bootstrap | 05:10 |
| wallyworld_ | arosales: we also need, it seems, to get other regions besides region a supported on hp cloud | 05:10 |
| arosales | wallyworld_, the cpc is going to be working on indexing hp in february | 05:10 |
| wallyworld_ | yay | 05:10 |
| jseutter_ | arosales: is there any secrets in the output? | 05:11 |
| arosales | jseutter_, nope | 05:11 |
| jseutter_ | arosales: http://pastebin.ubuntu.com/6841977/ | 05:12 |
| arosales | jseutter_, do you have | 05:14 |
| arosales | public-bucket-url: https://region-a.geo-1.objects.hpcloudsvc.com/v1/60502529753910 | 05:14 |
| arosales | for your hp stanza in your environments.yaml ? | 05:14 |
| wallyworld_ | arosales: public bucket not required | 05:16 |
| wallyworld_ | log shows: certified cloud tools-url set to https://region-a.geo-1.objects.hpcloudsvc.com:443/v1/60502529753910/juju-dist/tools | 05:17 |
| arosales | jseutter_, also here is an example of my hp stanza that I had working for 1.16 | 05:17 |
| arosales | wallyworld_, even for 1.16? | 05:17 |
| wallyworld_ | yeah | 05:17 |
| arosales | hmm | 05:17 |
| wallyworld_ | in 1.16, we made juju know about hp cloud | 05:17 |
| wallyworld_ | that line i just pasted above shows the tools url being set for hp cloud | 05:18 |
| arosales | wallyworld_, good point | 05:18 |
| wallyworld_ | this line shows tools meta data being read | 05:19 |
| arosales | wallyworld_, I had that from previous juju versions I guess | 05:19 |
| wallyworld_ | finding products at path "streams/v1/com.ubuntu.juju:released:tools.json" | 05:19 |
| wallyworld_ | yeah | 05:19 |
| wallyworld_ | this line shows tools being found | 05:19 |
| wallyworld_ | juju.environs.boostrap bootstrap.go:71 environs: picked newest version: 1.16.5 | 05:19 |
| jseutter_ | arosales: I added public-bucket-url, http://pastebin.ubuntu.com/6841998/ | 05:20 |
| wallyworld_ | last lines show that it's image metadata not found | 05:20 |
| wallyworld_ | on cloud-images.ubuntu.com | 05:20 |
| wallyworld_ | so tools are ok | 05:20 |
| arosales | jseutter_, same error? | 05:21 |
| wallyworld_ | jseutter_: public bucket url not needed, finding the image metadata is the issue | 05:21 |
| wallyworld_ | let me check the image metadata | 05:21 |
| jseutter_ | arosales: yeah, seems exactly the same | 05:21 |
| arosales | jseutter_, ah what is your default series? | 05:22 |
| wallyworld_ | arosales: jseutter_ the region is wrong | 05:22 |
| arosales | jseutter_, sorry to have your try public-bucket wallyworld_ pointed out the tools are being found | 05:22 |
| wallyworld_ | you need to set the region to az-1.region-a.geo-1 | 05:22 |
| wallyworld_ | not az1.region-a.geo-1 | 05:22 |
| arosales | wallyworld_, nice eye | 05:23 |
| jseutter_ | really? | 05:23 |
| * jseutter_ smacks forehead | 05:23 | |
| wallyworld_ | it took me squinting hard to see it :-) | 05:23 |
| arosales | jseutter_, make sure you have "region: az-1.region-a.geo-1" | 05:23 |
| wallyworld_ | so in summary: juju "just works" for hp cloud and region a. not urls in config needed | 05:24 |
| wallyworld_ | other regions coming in feb hopefully :-) | 05:24 |
| arosales | also of note we don't have saucy images in hp cloud yet . . just precise. However, it looks like you are using precise too based of your output | 05:24 |
| arosales | wallyworld_, thanks | 05:24 |
| wallyworld_ | np | 05:25 |
| jseutter_ | arosales: I don't have a default-series set for my hpcloud config section | 05:25 |
| arosales | jseutter_, no worries it should default to precise | 05:26 |
| arosales | and your output looks like that is what it is using | 05:26 |
| jseutter_ | arosales: error seems unchanged - http://pastebin.ubuntu.com/6842022/ | 05:27 |
| arosales | jseutter_, try to rm ~/.juju/environments/hp.jenv | 05:28 |
| arosales | given you call your environments.yaml "hp" | 05:28 |
| jseutter_ | arosales: I named it hpcloud, and I have been verifying that destroy-environment blows away the file in environments/ | 05:30 |
| jseutter_ | just tried it again, no change | 05:30 |
| jseutter_ | hm | 05:30 |
| wallyworld_ | jseutter_: it's az-1.region..... | 05:32 |
| wallyworld_ | not az-1-region..... | 05:32 |
| wallyworld_ | that info is not Juju btw, that comes from hp cloud | 05:33 |
| wallyworld_ | juju needs to match with how hp cloud exports its region names | 05:33 |
| jseutter_ | fsck, sorry about that | 05:33 |
| wallyworld_ | np :) | 05:33 |
| wallyworld_ | easy to do a typo for sure, i've made lots of similar mistakes myself | 05:33 |
| wallyworld_ | i also say a lot worse than fsck :-) | 05:34 |
| jseutter_ | finally, a different error: http://pastebin.ubuntu.com/6842041/ :) | 05:36 |
| jseutter_ | hehe, trying to keep it pc. Today I can verify that my keyboard is resistant to beatings | 05:36 |
| arosales | jseutter_, lol | 05:39 |
| wallyworld_ | jseutter_: i've not seen that one. juju got told to ask hp cloud to start an instance with image id 81078 | 05:39 |
| * arosales taking a look at your recent debug log | 05:39 | |
| wallyworld_ | hp cloud complained | 05:39 |
| wallyworld_ | perhaps the images ids we are using are out of date? | 05:40 |
| wallyworld_ | i think there's a nova command to list the current ids | 05:40 |
| wallyworld_ | nova image-list | 05:41 |
| jseutter_ | hm, is there some way to get a novarc out of hpcloud? | 05:41 |
| arosales | wallyworld_, but that would be metadata thought right? | 05:41 |
| arosales | s/thought/though/ | 05:41 |
| wallyworld_ | arosales: yeah, i've just done a listing and there's no such image id anymore | 05:41 |
| wallyworld_ | arosales: so our images metadata appears stake | 05:42 |
| wallyworld_ | stale | 05:42 |
| * arosales tries a bootstrap on hpcloud | 05:42 | |
| wallyworld_ | jseutter_: i just set up a novarc file with exported values for region, username, password etc | 05:43 |
| wallyworld_ | so you could do it my hand if you have all the values, which you do cause i'm guessing they are in your env.yaml | 05:43 |
| jseutter_ | wallyworld_: I'll give it a shot | 05:44 |
| arosales | ugh I am on 1.17 | 05:44 |
| arosales | jseutter_, https://docs.hpcloud.com/cli/nova | 05:45 |
| arosales | for the nova set up | 05:45 |
| wallyworld_ | jseutter_: here's mine http://pastebin.ubuntu.com/6842083/ | 05:47 |
| arosales | wallyworld_, so is image stream data different for 1.16 and 1.17 | 05:47 |
| wallyworld_ | arosales: no, should be the same. i'd say hp added new images and deleted old ones and we haven't kept our metadata current | 05:47 |
| arosales | wallyworld_, so I get a successful bootstrap on hpcloud with 1.17 | 05:47 |
| wallyworld_ | hmmm. do you know what image id was used? | 05:48 |
| wallyworld_ | what region? | 05:48 |
| wallyworld_ | az-1, 2 or 3? | 05:48 |
| wallyworld_ | arosales: the 3 image ids in our metadata are 49850, 68481, and 81078 | 05:50 |
| arosales | az1 | 05:50 |
| wallyworld_ | for regions 3, 2, 1 | 05:50 |
| arosales | wallyworld_, http://pastebin.ubuntu.com/6842103/ | 05:51 |
| wallyworld_ | arosales: can you look at the hp console and see what image id was chosen? | 05:52 |
| arosales | sure one sec | 05:52 |
| arosales | wallyworld_, 81078 | 05:55 |
| wallyworld_ | arosales: weird. that's the same image id as jseutter_ got but his request to hp cloud was rejected | 05:55 |
| wallyworld_ | maybe it was a transient error | 05:55 |
| arosales | the ids change per region | 05:56 |
| jseutter_ | hm, is there something else I need to click in the gui? | 05:56 |
| wallyworld_ | i think it was also region az-1 | 05:56 |
| arosales | jseutter_, are you in US West AZ 1 | 05:56 |
| arosales | jseutter_, are you in the classic or new console? | 05:58 |
| jseutter_ | arosales: I have "region: az-1.region-a.geo-1" in my config section | 05:58 |
| jseutter_ | arosales: new console | 05:58 |
| jseutter_ | I just signed up with a trial today, I didn't get a choice to use the old console | 05:58 |
| * arosales traversing their new console | 05:59 | |
| arosales | hmm the new console isn't showing me us west az1 compute image ids | 06:07 |
| arosales | jseutter_, I _think_ you are operating in a new region | 06:07 |
| jseutter_ | arosales: hm, that would make sense. Is there any way I can tell? When I launch an instance in the gui I still get az1 as a choice of where my instance will run | 06:08 |
| arosales | jseutter_, if you go here | 06:09 |
| arosales | http://www.hpcloud.com/console | 06:09 |
| arosales | do you see an option for HP Classic Cloud Management Console | 06:09 |
| jseutter_ | yes. I haven't used it yet | 06:10 |
| arosales | jseutter_, its different services from hp | 06:10 |
| arosales | pre and post 13.5 | 06:11 |
| jseutter_ | arosales: when I click the Activate button in the classic console, I get automagically booted to the new console | 06:11 |
| arosales | when I logged into the new console us west was not activated for me, even though I knew it was | 06:11 |
| arosales | jseutter_, interesting | 06:12 |
| arosales | jseutter_, on https://console.hpcloud.com/dashboard | 06:12 |
| arosales | do you see a button for us west az1 | 06:12 |
| arosales | under "infrastructure services" | 06:12 |
| jseutter_ | arosales: no. When I click anything on that page I get sent to horizon.hpcloud.com | 06:13 |
| arosales | hmm ok | 06:13 |
| arosales | jseutter_, out of curiosity does https://console.hpcloud.com/compute/az-1_region-a_geo-1/servers load for you | 06:14 |
| arosales | jseutter_, looking at the two consoles I don't see an Ubuntu 12.04 image under partner provided images in the new console | 06:15 |
| arosales | however I do see this image in the old console under partner provided images | 06:15 |
| jseutter_ | arosales: no. I get redirected to https://console.hpcloud.com/dashboard and there is a red error message in the top right that says "The requested service does not exist in the project" | 06:15 |
| arosales | and that is image 81078 which juju is trying to bootstrap | 06:15 |
| arosales | ok, this looks like an issue with hpcloud that they are not bringing forward partner provided images | 06:16 |
| jseutter_ | arosales: do you see this image? https://horizon.hpcloud.com/project/images_and_snapshots/8c096c29-a666-4b82-99c4-c77dc70cfb40/ | 06:17 |
| arosales | specifically they did not bring Ubuntu 12.04 (81078) that Canonical provided and that juju is trying to bootstrap | 06:17 |
| arosales | Warning: You are not authorized to access /project/images_and_snapshots/8c096c29-a666-4b82-99c4-c77dc70cfb40/ | 06:17 |
| jseutter_ | That for me is an image "Ubuntu Precise 12.04 LTS Server 64-bit 20121026 (b)" | 06:18 |
| jseutter_ | ah, I see that some of the images in the list say "Partner Provided", but the ubuntu ones don't | 06:20 |
| jseutter_ | err, "Partner Image: | 06:20 |
| arosales | jseutter_, basically there should be a 12.04 image under partner provided | 06:20 |
| jseutter_ | nuts. | 06:21 |
| arosales | jseutter_, since I have an account before I can still access that image in the pre 13.5 hp services | 06:21 |
| arosales | jseutter_, I am going to open a ticket with hp cloud | 06:21 |
| jseutter_ | and probably why no one saw it before now. | 06:21 |
| arosales | jseutter_, sorry won't help you right now but hopefully we should see a resolution soon | 06:21 |
| arosales | jseutter_, it may be that your account is new | 06:22 |
| jseutter_ | arosales: yeah, I just signed up a few hours ago | 06:22 |
| jseutter_ | arosales: thanks for all your help, I owe you and wallyworld_ a beer when I see you guys | 06:23 |
| arosales | jseutter_, nah thanks for the feedback. Its appreciated. | 06:23 |
| wallyworld_ | hey, my pleasure. pleased to help | 06:23 |
| wallyworld_ | sorry you had all the issues | 06:24 |
| jseutter_ | heh, we just added more work to the pile today :) | 06:24 |
| arosales | jseutter_, I'll give you a ping when I find a conclusion here and also update our juju docs | 06:24 |
| jseutter_ | arosales: cool, thanks | 06:25 |
| * arosales filing a hp support ticket | 06:25 | |
| arosales | jseutter_, there is always canonistack | 06:27 |
| jseutter_ | arosales: I've been fighting with it all day, haven't been able to successfully bootstrap _once_ today in either region. | 06:28 |
| jseutter_ | I tried devstack for a bit but gave up when I found an email saying it was a bad idea to try it with juju | 06:30 |
| arosales | jseutter_, geesh you just can't catch a break | 06:31 |
| jseutter_ | heh, it sorta feels that way. Thanks again | 06:32 |
| === zz_mwhudson is now known as mwhudson | ||
| * arosales confirmed that there is a bug on HPs end. Specifically HP had not moved 12.12 partner images over to 13.5 | 07:03 | |
| arosales | so new users can't see 12.12 services onlly 13.5 | 07:03 |
| arosales | thus bootstraping fails as the partner images are not in 13.5 :-( | 07:04 |
| arosales | hopefully there will be a resolution soon | 07:04 |
| arosales | wallyworld_, ^ fyi | 07:04 |
| wallyworld_ | ah ok, thaks | 07:04 |
| === CyberJacob|Away is now known as CyberJacob | ||
| ianous | Does anyone know if juju sends auth-token to retrieve files from buckets? | 08:11 |
| === mwhudson is now known as zz_mwhudson | ||
| === thumper-afk is now known as thumper | ||
| === BradCrittenden is now known as bac | ||
| === gary_poster|away is now known as gary_poster | ||
| tomixxx | hi, it seems i have some DNS resolve problems: when i enter ssh cloud1.master it says: "ssh: Could not resolve hostname cloud1.master: Name or service not known" | 13:39 |
| === natefinch is now known as natefinch-afk | ||
| lazyPower | tomixxx, that's not a valid FQDN. is cloud1.master the name of a service you've deployed with juju? | 13:47 |
| tomixxx | no, its the fqdn | 13:47 |
| tomixxx | ah sorry | 13:48 |
| tomixxx | de fqdn is "cloud1" | 13:48 |
| tomixxx | .master is added by MAAS because i have enabled "DHCP and DNS" | 13:48 |
| lazyPower | Whats the IP of cloud1? Can you try ssh'ing to the IP instead of the assigned hostname? | 13:49 |
| tomixxx | ip is 10.0.0.100 | 13:50 |
| tomixxx | and yes, i can ssh with that | 13:50 |
| tomixxx | but the problem is that juju tries to enter with "cloud1" and not with the IP so juju stucks when i call for example "juju status" | 13:50 |
| tomixxx | hmm it seems its sufficient to add the nameserver to the interface, namely "dns-nameservers 10.0.0.9" | 13:53 |
| marcoceppi | tomixxx: you need to set your resolv.conf on your maas master to point to your DNS server | 14:00 |
| marcoceppi | tomixxx: while this won't survive reboots, edit /etc/resolv.conf at the top add nameserver 10.0.0.1 save it then try to ssh to cloud1.master | 14:01 |
| tomixxx | hi marcoceppi: i have modified etc/interfaces and added a line "dns-nameserver " and now resolv.conf on restart is alrady set | 14:02 |
| tomixxx | i cann ssh now cloud1.master | 14:03 |
| marcoceppi | tomixxx: okay, cool. You'll need to edit you /etc/resolveconf/resolv.conf.d/head to have the nameserver 10.0.0.1 and on another line add search master | 14:07 |
| marcoceppi | then run `sudo resolveconf -u` and then all you should need to do is type ssh cloud1 (no need for .master anymore) and get to that box | 14:07 |
| marcoceppi | from there juju status, juju ssh, etc should work without issue and you can carry on deployting | 14:07 |
| tomixxx | i want to re-install juju complete because the dns problem was already there when i added the nodes to maas (when they failed to download some packages) | 14:20 |
| tomixxx | the problem is, i cannot delete cloud1.master, because the node is "allocated to root" because it is needed as juju bootstrap node. | 14:20 |
| marcoceppi | tomixxx: you should be able to run `juju destroy-environment maas` | 14:21 |
| marcoceppi | which should un-allocate it from root and return it to the pool of available machines | 14:21 |
| tomixxx | i have executed sudo apt-get purge juju before, so i cannot do it anymore :D | 14:21 |
| tomixxx | ok, i will reinstall juju package in order to call destroy-environemnt | 14:23 |
| tomixxx | damn does not work | 14:23 |
| tomixxx | it says "error: unrecognized args: ["maas"]" | 14:24 |
| tomixxx | kk, calling "juju destroy-environment" is sufficient | 14:26 |
| tomixxx | ah help... the nodes cannot download packages again... | 14:32 |
| tomixxx | can i have multiple "search" statements in resolv.conf? | 14:35 |
| tomixxx | ok, now ubuntu+juju is installed again on my first cloud node, but i stuck agian when i call "juju status" | 14:45 |
| tomixxx | is this ok, will juju status only work if installation of cloud node is finished? | 14:45 |
| tomixxx | at the moment, "ssh ubuntu@cloud1" does not work either | 14:46 |
| marcoceppi | tomixxx: yes, you need to wait for cloud1 to install | 14:51 |
| tomixxx | OK | 14:52 |
| tomixxx | ok, seems to work | 15:07 |
| tomixxx | juju status works | 15:07 |
| marcoceppi | tomixxx: then you're all set | 15:34 |
| tomixxx | we will see | 15:35 |
| tomixxx | however, i can deploy already services like mysql or rabbitmq-server :-) | 15:35 |
| === CyberJacob is now known as CyberJacob|Away | ||
| tomixxx | how can i check which services are already startet with juju? | 15:50 |
| lazyPower | tomixxx, juju status has a data point: agent-state | 15:54 |
| tomixxx | ahh i see | 15:55 |
| lazyPower | if agent-state = started, your charm has run its install hooks, and is in a ready state according to the orchestrator | 15:55 |
| tomixxx | a question: when i run juju-add-relation nova-compute rabbitmq-server | 15:56 |
| tomixxx | i got "ERROR ambiguous relation: "nova-compute rabbitmq-serveR" could refer to "novacompute:amqP rabbitmpq-server:ampq"; "nove-compute:nrpe-external-master rabbitmp-server:nrpe-external-master" | 15:57 |
| tomixxx | is that normal? | 15:57 |
| tomixxx | and some juju-services have angent-state: pending... | 15:58 |
| tomixxx | on which machine are these service installed? | 15:58 |
| tomixxx | i dont understand: i have only 2 nodes, and they are listet in "juju status" but also a few other nodes up to 8 nodes which are not listed... | 15:59 |
| tomixxx | Can someone look at this, please: http://pastebin.ubuntu.com/6844443 | 16:03 |
| tomixxx | why are there listed multiple nodes until node "8", when i only have two nodes? | 16:05 |
| lazyPower | tomixxx, that is normal, there are two interfaces define | 16:05 |
| lazyPower | *defined. And you need to be implicit in which relationship are you building | 16:05 |
| lazyPower | marcoceppi, tomixxx has some machine spin up issues it looks like. Any input on whats up with gomaasapi? | 16:07 |
| tomixxx | what do u mean with "two interfaces" ? | 16:07 |
| tomixxx | :( do i have destroyed sth? | 16:08 |
| tomixxx | the install guide says: 1.It's assumed the user has sufficient number of nodes available in MAAS. If you don't intend to use the helper application, juju-jitsu and its deploy-to option allowing you to co-locate services, then the minimum deployment outlined below requires a minimum of 10 machines including the juju bootstrap node and MAAS server. | 16:11 |
| tomixxx | Do i need really 10 PHYSICAL machines? | 16:12 |
| tomixxx | i have only 2 | 16:12 |
| tomixxx | 3 together with maas | 16:12 |
| tomixxx | i want to start all service on one physical machine i guess | 16:14 |
| tomixxx | lazyPower: can i simply ignore the nodes 2 to 8 ? | 16:19 |
| tomixxx | lazyPower: Do you mean they come from the university network? | 16:19 |
| lazyPower | tomixxx, 1 moment | 16:19 |
| tomixxx | kk | 16:19 |
| lazyPower | tomixxx, ah i'm not sure about the requirement of 10 physical machines. | 16:23 |
| lazyPower | My exposure to maas has been very limited | 16:23 |
| marcoceppi | tomixxx: what are you trying to do at the end of the day? | 16:24 |
| marcoceppi | you do not need juju-jitsu anymore, --to support is in juju now | 16:24 |
| tomixxx | marcoceppi: at the end of the day, i want to parallelize a Natural-Language-processing task with cloud computing | 16:25 |
| tomixxx | so, my vision is that i can use openstack which will distribute to work to the nodes available in the cloud, roughly spoken | 16:26 |
| marcoceppi | tomixxx: so, if you want to deploy openstack, you can do a lot of it in containers, which will allow you to run more density on fewer nodes | 16:26 |
| marcoceppi | I think the least I've seen is something like 4 nodes for openstack. Then each additional node is just used for nova-computer | 16:27 |
| tomixxx | additionally, the performance of this private cloud is going to be evaluated against the performance of public cloud solutions | 16:28 |
| tomixxx | my question is now: can i deploy all service listed here: https://help.ubuntu.com/community/UbuntuCloudInfrastructure on one single machine? | 16:30 |
| tomixxx | and the second quesiton is: can i ignore the listed nodes 2 to 8 in the juju status output | 16:30 |
| marcoceppi | tomixxx: what do you mean the listed nodes 2 to 8? | 16:31 |
| tomixxx | http://pastebin.ubuntu.com/6844443 here u can see the output from "juju status" | 16:31 |
| tomixxx | i have only 2 physical nodes, but other nodes are listet there, however, indicating, they are not really available | 16:31 |
| marcoceppi | tomixxx: no, those mean that juju can't allocate those nodes | 16:32 |
| marcoceppi | tomixxx: you only have two MAAS nodes? | 16:32 |
| tomixxx | yes | 16:32 |
| marcoceppi | tomixxx: yeah, so that's why. Default juju story is to put one machine on one node | 16:32 |
| marcoceppi | tomixxx: it's hard to demo a cloud infrastructure you dont' have a lot of cloud space | 16:32 |
| tomixxx | ;-) | 16:33 |
| marcoceppi | tomixxx: You can put almost all the services, execept nova-compute, on LXC containers on one node | 16:33 |
| marcoceppi | tomixxx: if you don't use swift, or any object store, you /could/ use just two machines | 16:33 |
| marcoceppi | tomixxx: so what you'll want to do, is juju destroy-environment | 16:34 |
| marcoceppi | this will re-allocate the nodes back to maas | 16:34 |
| tomixxx | k | 16:35 |
| tomixxx | done | 16:35 |
| marcoceppi | tomixxx: give me about 5 mins | 16:36 |
| tomixxx | k, the whole task looks as follow: a user uploads a number of plank text documents to the cloud, the cloud distributes the documents to the nodes (i.e. with map reduce or similar techniques), each node make the languaga analysis and store the results in a distributed file system and finally, the user is able to download the results from the cloud. | 16:37 |
| tomixxx | so, roughly spoken, i need to get things work simply because i need to write my master thesis and maybe i will use this technology later on in my company as well. | 16:38 |
| tomixxx | so, i do this not "for fun" | 16:38 |
| tomixxx | marcoceppi: are u still here? | 16:54 |
| marcoceppi | tomixxx: yes, sorry | 16:55 |
| tomixxx | np | 16:55 |
| marcoceppi | tomixxx: you can use juju deploy --to lxc:0 for almost all the services except neutron (so don't use neutron for networking) and nova-compute | 16:56 |
| marcoceppi | when you get to deploying nova-compute, just run juju deploy nova-compute to give it it's own machine | 16:56 |
| tomixxx | kk, why do nova-compute need its own node? | 16:57 |
| tomixxx | (just for the sake of interest) | 16:57 |
| marcoceppi | now this may not work for openstack, but almost all the services can be put in LXC containers on one node except nova-compute which needs it's own node | 16:57 |
| marcoceppi | tomixxx: because that's what actually spins up VMS | 16:57 |
| tomixxx | k | 16:57 |
| marcoceppi | tomixxx: so when you want more compacity, you add another machine in maas, then you can juju add-unit nova-compute and you have more resources to run VMs in openstack | 16:57 |
| tomixxx | k | 16:58 |
| marcoceppi | you also can't use ceph, swift, or any of the osd stuff because of the lxc containers, you need physical machines for that | 16:58 |
| marcoceppi | but if you don' need an object store, just the VM capacity, that should do | 16:58 |
| tomixxx | what do u mean with "object store" | 16:58 |
| === natefinch-afk is now known as natefinch | ||
| tomixxx | marcoceppi: do i need to bootstrap juju again? | 17:00 |
| marcoceppi | tomixxx: yes, after destroy-environment you'll need to bootstrap | 17:01 |
| tomixxx | ok, i did "juju bootstrap --upload-tools" and it says " ERROR TLS handshake failed: x509: certificate sigend by unknown authority" | 17:01 |
| marcoceppi | tomixxx: that's interesting. run `rm -f ~/.juju/environments/*.jenv` then try again | 17:03 |
| tomixxx | and do i need to reboot the node i guess? | 17:06 |
| tomixxx | on which juju is going to be deployed | 17:07 |
| tomixxx | because "juju status" stucks now again | 17:07 |
| tomixxx | because i have no wake-on-lan ;) | 17:09 |
| tomixxx | marcoceppi: however, i guess i know how to continue now and I thank you for your help so far ;) | 17:12 |
| marcoceppi | tomixxx: yes, you'll need to reboot it | 17:13 |
| tomixxx | kk | 17:13 |
| mgz | what on on unholy venv have you done to the amulet test setup marcoceppi... | 18:26 |
| marcoceppi | mgz: not me | 18:26 |
| marcoceppi | blame bcsaller and rick_h_ ;) | 18:27 |
| rick_h_ | :) | 18:27 |
| mgz | it doesn't even work, skips the site-packages but doesn't include python3-yaml in the test deps | 18:27 |
| marcoceppi | mgz: there's a lock step thing, or something, you have to do | 18:27 |
| marcoceppi | like make build; make venv; make test | 18:28 |
| marcoceppi | I think | 18:28 |
| * marcoceppi looks | 18:28 | |
| rick_h_ | make sysdeps && make && make test | 18:28 |
| rick_h_ | I think | 18:28 |
| rick_h_ | the whole thing stinks, but some issue in trust vs older and the path the py3 venv puts things (bin vs local/bin) | 18:29 |
| rick_h_ | trusty that is | 18:29 |
| * marcoceppi confirms | 18:29 | |
| mgz | okay, now it ImportErrors on "from . import helpers" in amulet.deployer | 18:29 |
| marcoceppi | mgz: if I run make sysdep && make && make test | 18:30 |
| rick_h_ | that I know nothing about | 18:30 |
| marcoceppi | the testing suite runs | 18:30 |
| * rick_h_ goes to check it out | 18:30 | |
| mgz | looks like it lacks charmworldlib | 18:30 |
| marcoceppi | and succeds | 18:30 |
| marcoceppi | mgz: is it because something is broken in helpers? | 18:31 |
| marcoceppi | did you update it at all? | 18:31 |
| mgz | no, just charmworldlib not existing | 18:31 |
| * marcoceppi scratches head, sounds like venv isn' working or something | 18:32 | |
| mgz | there's an egg in the venv... | 18:32 |
| mgz | er... I need to remember how to do some of this | 18:32 |
| rick_h_ | mgz: activate the venv? | 18:32 |
| rick_h_ | mgz: is this running the tests? they should auto be in the venv paths | 18:33 |
| mgz | rick_h_: trying to diagnose why the tests fail | 18:33 |
| mgz | import errors are tricky, particulary when I can't trivially reproduce the environment the import is happening in | 18:33 |
| rick_h_ | mgz: k, building here with the latest trunk to see if I can dupe | 18:34 |
| rick_h_ | mgz: the import is from the test then? | 18:34 |
| rick_h_ | marcoceppi: we should update to use the package name for imports vs the . as well imo. | 18:34 |
| mgz | yeah, just need to get my test stuff working, venv/bin/python3 helps | 18:34 |
| marcoceppi | rick_h_: ack | 18:35 |
| mgz | rick_h_: generally an import failing like that means some *other* import failed, I have it now | 18:35 |
| rick_h_ | mgz: k, tests pass trunk here. not seeing the error. It's all virtualenv'd so if you run things manually you need to either activate it or work through venv/bin to do things | 18:35 |
| mgz | accidentally wrote a line of python in go :) | 18:35 |
| rick_h_ | lol, did it get cranky? | 18:36 |
| marcoceppi | ;) | 18:36 |
| rick_h_ | mgz: suggestions for improvement welceom. The goal is to get tox and multiple python versions going which is part of the reason for the process. | 18:38 |
| === CyberJacob|Away is now known as CyberJacob | ||
| marcoceppi | hazmat: you around | 19:08 |
| marcoceppi | ? | 19:08 |
| hazmat | marcoceppi, yes | 19:12 |
| marcoceppi | hazmat: sorry, false alarm, was maas tom foolery | 19:12 |
| hazmat | rick_h_, is anyone working on making charmworld ui / manage.jujucharms.com look decent/functional? | 19:12 |
| rick_h_ | hazmat: not atm. | 19:13 |
| rick_h_ | hazmat: it's pending some project talk next week and not currently on any project plan I know of atm. | 19:13 |
| hazmat | rick_h_, ic, thanks | 19:13 |
| rick_h_ | hazmat: but jcastro has a giant cheering flag he waves around the topic :) | 19:14 |
| jcastro | I have an item on the spreadsheet for the sprint | 19:14 |
| hazmat | rick_h_, its painful.. to use.. but its the only thing to use.. i get tempted to resurrect the old code base on a machine just to have a compact usable ui. | 19:14 |
| jcastro | hazmat, let's determine that next week | 19:15 |
| jcastro | I hear ya | 19:15 |
| jcastro | but like, we have an existing mess and adding another UI would make it worse imo | 19:15 |
| rick_h_ | hazmat: write a cli client using the api for us :) /me has pondered that a couple of times | 19:16 |
| hazmat | rick_h_, well.. part of the issue is that api is in the wrong place.. i'm not sure promoting it more is the right answer.. and really a cli isn't going to help for information display. | 19:17 |
| rick_h_ | hazmat: +1 honestly I'm going to be all on board with the planned session of splitting the store out of the juju base and building it up. | 19:18 |
| === zz_mwhudson is now known as mwhudson | ||
| === thumper is now known as thumper-afk | ||
| avoine | someone have tried to send json with the relation-set command? | 20:59 |
| marcoceppi | avoine: the relation-set command can take json with --format json | 21:01 |
| avoine | cool, I'll use that | 21:02 |
| === thumper-afk is now known as thumper | ||
| jamespage | marcoceppi, https://launchpad.net/~mysql-ubuntu/+archive/mysql-5.6/+packages | 21:54 |
| jamespage | first cut of mysql-5.6 package for 14.04 | 21:54 |
| === thumper is now known as thumper-afk | ||
| marcoceppi | mgz: thank you for your pr, we found the issue with bzr. Apparently it really want's to know whoami | 22:57 |
| marcoceppi | jamespage: thank you sir! | 22:57 |
| jamespage | marcoceppi, np | 22:57 |
| === CyberJacob is now known as CyberJacob|Away | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!