| === frankban|afk is now known as frankban | ||
| kjackal | Good morning Juju world! | 08:06 |
|---|---|---|
| BlackDex | Hello there, is there still an option to force the usage of LXC instead of LXD with juju 2.0? | 08:27 |
| marcoceppi | BlackDex: lxd is lxc, you mean lxc 1.0 instead of 2.0? | 08:27 |
| BlackDex | yes.. i think i mean that... | 08:28 |
| BlackDex | Am i correct that LXD an KVM/QEMU can't run together | 08:28 |
| BlackDex | Because of the CPU VT-D usage? | 08:29 |
| rockstar | BlackDex: LXD and KVM can absolutely be used together. I'm doing it on my machine right now. | 08:39 |
| rockstar | BlackDex: I can't imagine a reason why LXD would be using VT at all. It's not a virtualized environment. | 08:40 |
| BlackDex | Ah oke | 08:46 |
| BlackDex | So, if i want to deploy openstack using juju 2.0 and run some services on a compute node that can be done withing LXD :) | 08:46 |
| marcoceppi | BlackDex: yeah, you should check out: https://github.com/openstack-charmers/openstack-on-lxd | 08:49 |
| rockstar | BlackDex: absolutely. | 08:49 |
| rockstar | BlackDex: in fact, I do the development of nova-lxd inside a lxd instance, so my devstack compute nodes are straight-up lxdception. | 08:49 |
| BlackDex | rockstar: So you have a Host, which has an LXD container, which has nova-lxd in it? | 08:51 |
| rockstar | BlackDex: yes. | 08:51 |
| BlackDex | Wow | 08:51 |
| BlackDex | nice | 08:51 |
| rockstar | And I've also booted KVM instances inside of my nova-lxd container. | 08:51 |
| rockstar | BlackDex: it's changed a bit since this, but if you're interested: http://iamtherockstar.com/devstack-in-a-lxd-container.html | 08:51 |
| BlackDex | including VT? | 08:51 |
| rockstar | BlackDex: yes. | 08:55 |
| BlackDex | Oke, nice | 08:56 |
| BlackDex | thx for the info! and the website | 08:56 |
| rockstar | BlackDex: LXD shares the kernel with the host, so as long as the kernel has support for it, the LXD gets them too. | 09:02 |
| BlackDex | oke, that is good to know. | 09:02 |
| BlackDex | Do you also know if it is possible to have static ip's for LXD/LXC containers via JuJu/Maas? | 09:03 |
| BlackDex | Can't seem to find anything on this. It all seems to do DHCP | 09:03 |
| anrah | Hi! I upgraded my OpenStack and during the update I enabled HTTPS on my public APIs | 09:10 |
| anrah | now I get error: http://paste.ubuntu.com/23424696/ | 09:11 |
| anrah | the instance gets created but apparentely the bootstrapped machine needs the ca-file to the machine if it wants to talk to openstack | 09:11 |
| anrah | ca-cert-path config option should do the trick, but that is not available on juju 2.0 ? | 09:12 |
| marcoceppi | anrah: it might be best to check in #openstack-charms | 09:14 |
| rockstar | BlackDex: that'd be a question for #lxcontainers | 09:16 |
| BlackDex | rockstar: Thx i will check over there | 09:16 |
| caribou | Hi, is there a way to prevent amulet from destroying the environment after tests (juju 1.25) ? | 09:17 |
| caribou | for some reason rsync refuses to get my tests logs; I get permission denied | 09:18 |
| anrah | marcoceppi: I am not deploying openstack, but rather definining openstack as a cloud which juju should use | 09:18 |
| kjackal | caribou: are you using bundletester to run your tests? | 09:24 |
| caribou | kjackal: no, just "juju test" from the charm's directory | 09:24 |
| kjackal | caribou: could you try the reset: false flag inside the test.yaml? | 09:25 |
| kjackal | caribou: I am not sure if that will be honored. Have a loog at the description of the flag here: https://github.com/juju-solutions/bundletester | 09:26 |
| caribou | kjackal: test.yaml ? don't have one of those, just the test scripts | 09:26 |
| caribou | kjackal: ok, I'll look at it | 09:26 |
| caribou | kjackal: thanks! | 09:31 |
| cory_fu | kjackal: Can you give me a review on https://github.com/juju-solutions/cloud-weather-report/pull/81 | 10:15 |
| kjackal | cory_fu: I am on it! | 10:15 |
| cory_fu | Thanks! | 10:15 |
| kjackal | cory_fu: Looks good! Let me know if you want me to merge it. Should we wait for the 200 amulet return? | 10:30 |
| cory_fu | kjackal: Were you able to run it and confirm the HTML fix? | 10:31 |
| kjackal | Oh, you need a full use then! Ok, that will take some more time. I just run the unit test. I will ping you again | 10:33 |
| kjackal | cory_fu: ^ | 10:33 |
| cory_fu | Please do. Thank you | 10:33 |
| cory_fu | kjackal: Heading to lunch now. bbiab | 10:33 |
| kjackal | cory_fu: run the testplan (git) exmaple on lxd. No problems in the HTML, looks fine! | 10:46 |
| cory_fu | kjackal: great, thanks | 11:09 |
| === linaras is now known as deanman | ||
| petevg | bcsaller, cory_fu: I just pushed an update to https://github.com/juju-solutions/matrix/pull/11, which fixes the issues that I was seeing in test_glitch. You may need to rebuild your .tox environment, as it incorporates a matching fix to python-libjuju | 15:19 |
| BlackDex | can i mix series within a bundle file? | 15:52 |
| aisrael | blackboxsw: Yes | 15:52 |
| BlackDex | like almost everything is xenial, but some stuff is trusty, and that will be in an lxc? | 15:52 |
| BlackDex | the default series can be xenial, and then a charm just trusty/something, and that will work? | 15:53 |
| aisrael | blackboxsw: here's an example: http://pastebin.ubuntu.com/23426074/ | 15:53 |
| aisrael | You can set a per-machine series | 15:54 |
| blackboxsw | BlackDex, that was for you ;) | 15:54 |
| aisrael | err, yes, sorry! | 15:54 |
| BlackDex | :) | 15:55 |
| aisrael | silly tab completion | 15:55 |
| BlackDex | hehe, but i figured :) | 15:55 |
| BlackDex | oke thx, ill go and try that | 15:55 |
| jjohnston | Happy Friday, I have a question about juju 2.0 and ipv6 if anyone here can assist? My question is how to I prevent lxd instances from taking ipv6 addresses during a juju deploy? | 17:33 |
| jrwren | jjohnston: using the local lxd provider? | 17:34 |
| jjohnston | no, deploying to remote nodes via maas | 17:35 |
| jjohnston | maas managed nodes rather | 17:35 |
| jjohnston | i don't have any ipv6 networks in maas and the bare-metal hosts take ipv4 addresses as expected, but the lxd instances, including the api endpoint for ceph (in this case) are taking ipv6 addresses | 17:37 |
| === frankban is now known as frankban|afk | ||
| marcoceppi | jjohnston: thta's really odd, I htink there's a model config to disable ipv6 | 17:42 |
| * marcoceppi checks | 17:42 | |
| jjohnston | that would be nice, I'm green with juju, but i've not been able to find such a parameter after extensive googling | 17:43 |
| marcoceppi | jjohnston: I don't have a MAAS model available to me atm, but does juju model-config yield anything exciting? | 17:43 |
| rock | Hi all. To push our charm to charm public store we created an charm account using https://login.ubuntu.com/xbX4GvuWSuVkirqs/+decide. I pushed charm to store using that account from JUJU CLI. Now I deleted my charm account permanently. Bur account user name is still there. | 17:43 |
| rock | How can I remove that account username | 17:44 |
| marcoceppi | rock: which account username? | 17:44 |
| rock | marcoceppi: Hi. ubuntu SSO account username. | 17:45 |
| marcoceppi | rock: right, what is your username you want to delete? :) | 17:46 |
| rock | marcoceppi: "kaminario" | 17:46 |
| rock | When I was trying to create another account using same username "kaminario". It was showing username "kaminario" already in use. | 17:48 |
| marcoceppi | rock: I will take a look | 17:50 |
| rock | marcoceppi: Thank you. | 17:50 |
| jjohnston | @marcoceppi the only thing of remote interest that I can see is the ignore-machin-addresses and that's only because I can't even guess what that does | 18:15 |
| jjohnston | i don't see anything regarding protocol preferences as previous versions of juju apparently had | 18:16 |
| marcoceppi | jjohnston: this might be a better topic for juju@lists.ubuntu.com and juju-dev@lists.ubuntu.com, I don't have a good answer for you | 18:55 |
| pragsmike | anyone else having trouble rabbitmq charm rev 56 to run in openstack-base? | 19:02 |
| pragsmike | all the openstack charms, mysql, ntp started ok, just rabbitmq says error, hook failed: "config-changed" | 19:03 |
| pragsmike | deploying using maas 2.1 | 19:04 |
| marcoceppi | pragsmike: are you putting rabbit in a lxd machine? | 19:12 |
| jjohnston | @marcoceppi: Thanks, I'm going to redploy juju ensuring there are no references to ipv6 prior to the bootstrap | 19:14 |
| marcoceppi | jjohnston: cool, if maas isn't configuring ipv6, it really shouldn't get addresses at the contiainer level | 19:15 |
| marcoceppi | fwiw, I haven't encountered that yet | 19:15 |
| pragsmike | marcoceppi: yes, I'm using the openstack-base bundle (except I tweaked constraints on the machines section) | 19:16 |
| pragsmike | marcoceppi: and that bundle does specify that rabbitmq be in lxc | 19:17 |
| marcoceppi | pragsmike: right, I wonder if there's a fix out | 19:17 |
| pragsmike | is that what the problem is? | 19:19 |
| marcoceppi | nope, that's the lastest | 19:19 |
| marcoceppi | beisner: have you come across this ^ | 19:20 |
| marcoceppi | coreycb: ^ | 19:20 |
| pragsmike | I did change the bundle to use rabbitmq rev 56, where it was 54 | 19:21 |
| pragsmike | but i did that because it was failing the same way before | 19:22 |
| marcoceppi | pragsmike: ah, excellent. I know rabbitmq-server freaks out when things like DNS don't work | 19:23 |
| marcoceppi | you're on maas 2.0 or 2.1? | 19:23 |
| pragsmike | maas 2.1 | 19:24 |
| marcoceppi | pragsmike: also, can you pastebin the /var/log/juju/unit-rabbitmq-server*.log file | 19:24 |
| pragsmike | marcoceppi: http://pastebin.com/v2XYtbzM | 19:29 |
| pragsmike | that just seems to be the rabbitmq client failing to connect to the server, doesn't say why the server didn't start | 19:30 |
| pragsmike | also I'm not sure if it's ok that it thinks the host name is "eth3". maas has decided that address has the name eth3.juju-8566c3-0-lxd-2 | 19:35 |
| pragsmike | I have a bunch of networks set up per Dimiter's instructions, but I haven't tried to make anything use them yet. juju just decided to use the 10.100 network for everything, which ought to be ok to start. | 19:37 |
| pragsmike | rabbitmq doesn't use ceph, does it? | 19:38 |
| kwmonroe | petevg: would you mind taking a look? https://github.com/juju-solutions/layer-apache-bigtop-base/pull/53 | 19:44 |
| petevg | kwmonroe: looking ... | 19:44 |
| petevg | kwmonroe: left some comments. | 19:49 |
| pragsmike | marcoceppi: also http://pastebin.com/D0NPZ4xJ is dmesg from rabbitmq container, full of segfaults in beam.smp | 19:55 |
| pragsmike | is remove-application suuposed to work in 2.0? juju remove-application rabbitmq-server has no effect that I can see. | 20:33 |
| === zz_CyberJacob is now known as CyberJacob | ||
| kwmonroe | sorry to make you grump petevg -- thx for the review ;) | 21:15 |
| petevg | kwmonroe: no problem :-) | 21:16 |
| petevg | kwmonroe: squashed n merged. | 21:17 |
| kwmonroe | you da man. | 21:17 |
| coreycb | thedac, pragsmike had an issue with rmq earlier. i think you had a fix queued up, i wonder if it's related. | 21:30 |
| thedac | My rabbit fix is related to maas 2.0 DNS and multiple interfaces | 21:31 |
| * thedac reads backscroll | 21:31 | |
| thedac | pragsmike: yes, I think my change will help you cs:~thedac/rabbitmq-server-3 https://review.openstack.org/#/c/389387/ for reference | 21:33 |
| pragsmike | ok thanks guys, I'll try that one | 21:48 |
| pragsmike | but once the bundle has deployed, juju seems to ignore any attempts i make to remove applications | 21:48 |
| pragsmike | thedac: what's the recommended way to replace a running rabbitmq app with yours? | 21:49 |
| thedac | pragsmike: you could do `juju upgrade-charm rabbitmq-server --switch cs:~thedac/rabbitmq-server-3` | 21:49 |
| thedac | Actually probably need a --force in there too | 21:50 |
| thedac | Then a juju resolved rabbitmq-server to get it going again. | 21:50 |
| pragsmike | k thanks | 21:50 |
| pragsmike | thedac: hmm, same problem, on the surface | 21:54 |
| pragsmike | let me investigate further | 21:54 |
| pragsmike | maybe redeploy the whole shebang with that new charm | 21:55 |
| thedac | pragsmike: hmm, I wonder if /etc/rabbitmq/rabbitmq-env.conf still has NODNAME=eht3? That is the problem | 21:55 |
| thedac | It might take a bit of manual intervention | 21:55 |
| pragsmike | i will look | 21:55 |
| thedac | A redeploy would fix it :) | 21:55 |
| pragsmike | it does have RABBITMQ_NODENAME=rabbit@eth3 | 21:56 |
| thedac | So just some backstory. rabbitmq is very sensative to DNS. It also needs a shortname for NODENAME. and eth3 is not resolvable. So a quick hack is to add eth3 to the 127.0.0.1 line in /etc/hosts | 21:57 |
| thedac | my change completely avoids this problem by not setting NODNAME and forceing `hostname` rather than doing DNS reverse lookup of the iP | 21:58 |
| pragsmike | i just commented it out and did "resolved". Aha! Now it's saying "(config-changed) RabbitMQ failed to start" in status, which i never saw before | 21:58 |
| pragsmike | ah, now it's back to "config change failed" | 21:59 |
| pragsmike | I'm going to just redeploy and go to dinner | 21:59 |
| thedac | pragsmike: I suspect rabbit is a bit hosed at this point. | 21:59 |
| thedac | Yes that is the quicker solution | 21:59 |
| pragsmike | thanks for the quick help and explanation | 21:59 |
| thedac | no problem | 21:59 |
| pragsmike | Recovery-oriented Computing | 21:59 |
| pragsmike | it's a thing | 21:59 |
| thedac | I'll be getting that change merged next week. | 21:59 |
| thedac | Rabbitmq is the most fragile service in the OpenStack | 22:00 |
| pragsmike | wow that's saying something | 22:00 |
| thedac | :) | 22:00 |
| pragsmike | i'd have voted for heat | 22:00 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!