/srv/irclogs.ubuntu.com/2017/05/12/#juju.txt

erik_lonroth3Good moring jujuers06:29
=== frankban|afk is now known as frankban
kjackalGood morning Juju world!09:00
kjackalthank you for taking the time to write this email erik_lonroth309:01
Zichi here, iirc, the etcd charm nos do daily backup on its own, do you know where?12:06
Zics/nos/now/12:06
=== scuttlemonkey is now known as scuttle|afk
lazyPowerZic: Ok clue me in as to where you got that info13:13
lazyPowerbecause thats news to me :)13:13
Zicsounds like I'm wrong here xD13:15
Zicoh I remembered: etcd do an automatical backup when upgrading from .deb to snap, but there is no "daily" ones, I just dreamed about13:16
lazyPowerZic: right, you can also snapshot using the action... which you could put on a cron job to run the action, parse the action output and fetch teh snapshot to do dailies13:16
lazyPowerthe primitives are all there to do it... so there's nothing really stopping you from doing that as a jenkins job13:17
Zicyup, it will be sexier than my old cron13:18
Zicwhich looks like this today: 0 0 * * * root cd /data/etcd-backup && etcdctl backup -data-dir=/var/lib/etcd/default -backup-dir=etcd-backup_$(date -I) && tar zcf etcd-backup_$(date -I).tar.gz etcd-backup_$(date -I) && rm -rf etcd-backup_$(date -I)13:18
Zic(I got to update this one with the new path of "etcdctl" as "/snap/bin/etcdctl")13:19
lazyPower:) I have an open todo to change the snapshot format to support the etcdctl backup command13:19
lazyPowerright now it just tarballs up the data directory, when you reinit the cluster from the snapshot it takes care of cleaning up any dirty state that may be left around in the db13:20
* Zic add to his TODO "Test to restore and etcd backup in case of disaster"13:22
Zicnever did it, and as you all know, you can't call something a "backup" since you don't have test it to restore13:22
Zic:>13:22
ZiclazyPower: what is the complete line to run the action ? I have a missing parameter :o13:32
lazyPowerZic: juju actions etcd --format=yaml --schema13:33
Zicthx13:35
ZiclazyPower: oh, I forget to tell you that the main CDK cluster (the big one) is now publicly accessible, the Android & iOS app has been released by our customer yesterday (as beta) on mobile stores13:36
lazyPower\o/13:37
lazyPowerwe made it13:37
Zic:crossfinger:13:37
lazyPowerthats awesome that you're in that new milestone of the journey13:37
ZiclazyPower: hmm, I have a strange thing with juju run-action etcd/0 snapshot14:07
Zichttp://paste.ubuntu.com/24560810/14:07
Zicit show "exited status 1" in show-action-status14:07
Zicsee above for juju debug-log14:07
lazyPowerhmm14:08
lazyPowerok, can you bug that for me Zic and I'll take a deeper look? i'm not certain off the top of my head14:08
Zic"open /var/snap/etcd/current/member/snap: no such file or directory" <= the correct path is /var/snap/etcd/current/etcd2.etcd/member/snap14:09
lazyPowerah14:09
lazyPowerthe migration path schenanigans14:09
lazyPowerI -completely- forgot about that14:09
lazyPowergood catch Zic, that should be a simple fix, i just need to do a quick path check for unit_name as teh first data path in the backup action before presuming its the default location.14:10
Zicdo you want that I fill a bug in the bundle-canonical-kubernetes@GitHub?14:10
lazyPowercertainly. i can cross post to the layer-etcd repo and get both when i push a fix later14:11
ZiclazyPower: https://github.com/juju-solutions/bundle-canonical-kubernetes/issues/28514:16
lazyPowerZic: thanks for filing that and sorry about the bug D:  I'll make sure that a fix for that makes its way to the layer today14:23
Zicno problem :) was just searching if we always need our home ugly (= as behaviour can change between version) cron or if I can switch to something more standard handled by Juju14:25
Zicwe skip a few backup days of etcd the time I saw that the old cron does not work anymore since we siwched to snap14:25
Zic+w14:25
lazyPowerZic: yeah, i'd like to keep the admin interface clean for you as well14:52
lazyPowerZic: feel free to file any usability bugs that you feel would make your life easier14:52
rahworkxHello all, how can I specify apt cache when using conjur-up15:03
lazyPowerrahworkx: as far as i know you would have to add the model, and set it on the model-config, then use conjure-up in headless mode15:26
lazyPowercory_fu: does that sound correct?15:26
lazyPowerooo wait, what about model-defaults?15:26
cory_fustokachu: ^15:26
stokachurahworkx: conjure-up -h15:27
cory_fulazyPower: Yeah, you could probably manually bootstrap a controller, set some model defaults, and then select that controller with conjure-up15:27
stokachuapt-proxy and apt-https-proxy are available15:27
cory_fuOh, look at that15:27
cory_fu:)15:27
rahworkxok thanks will take a look.15:33
stokachurahworkx: thanks, feel free to ping in here if you need help15:43
rahworkxstokachu: thanks will do.15:44
=== frankban is now known as frankban|afk
Budgie^SmorelazyPower want to see the case design idea I have for a cluster in a box?18:32
Budgie^Smorehttps://goo.gl/photos/V8UxiobqgNA6Xtn3918:33
bdxBudgie^Smore: want to see mine?18:46
bdxhttp://imgur.com/a/BYHPG18:46
bdxhttp://imgur.com/a/DpsrW18:46
bdxhttp://imgur.com/a/GDpyO18:47
bdxshe has been through many stages18:48
bdxhttp://imgur.com/a/SU1vd18:48
bdxhttp://imgur.com/a/oUQEU18:50
bdxhttp://imgur.com/a/fYeWV18:53
Budgie^Smoreoh nice a nvidia tesla!... I have been meaning to build a desk case for my main workstation and go with liquid cooling... the box in that link is for a 5 node mini itx with a 16 port switch18:54
bdxoh sick18:54
Budgie^Smorewas thinking individual aio liquid cooled cpu coolers for it18:54
Budgie^Smoreat least for the "worker" nodes18:54
jrwrenthat is a lot of GPU power.18:54
Budgie^Smorewell it would only be the onboad GPUs... not really looking at GPU power as I didn't design in space for a PCIe card18:55
Budgie^Smoreoh the tesla, yeah that is a nice card :)18:56
bdxI rip the resistors off the gtx cards and turn them into grid and teslas18:57
bdxsuper simple18:57
Budgie^Smoredid you get the case custom made?, just looking at the drive bay layout18:57
bdxyeah18:58
bdxunfortunately danger den closed their doors a few months after I ordered that18:58
Budgie^Smoreyeah I have heard of them :-/18:58
Budgie^SmoreI would really like to figure out how to get customer boards built so I could basically build a midplane with connectors for multiple systemboards, etc. and make the unit a lot smaller19:00
bdxBudgie^Smore: have you considered just doing a microblade enclosure?19:03
bdxin a mini rack19:03
bdxhttp://www.wiredzone.com/startech-racks-kvm-chassis-power-racks-and-cabinets-enclosure-cabinets-2636cabinet-3095624219:04
lazyPowerwow Budgie^Smore nice subwoofer box! "Thats my computer...." Nice computer! ;D19:05
lazyPoweri tease, this looks neat19:05
lazyPowerbdx: thats a nice lookin rig too19:06
lazyPowerBudgie^Smore: whats budget on that box build looking like?19:06
bdxlazyPower: thx19:07
Budgie^Smoreyeah just do you want to spend 1k on a glorified case? that is also massive in comparison... my design is basically 22.25" x 16.5" x 12"19:08
Budgie^SmorelazyPower for the parts you are talking 5 x itx board (recommend one with ipmi like the asrock on at $210 each) but you could use a small speced board for the maas node19:09
shann_Hi, juju purpose endpoint for charms inside hypervisor ? i try find documentation for subject19:09
Budgie^Smorethe PSUs I found that are a nice form factor are $30 each19:09
lazyPowershann_: I'm not sure i understood the question. Are you looking for the network endpoint of the controller node?19:10
Budgie^Smoreif you want liquid cooled it is another $70 / node and the case can handle 4 of those / 2 layers, but easily use some 120mm fans and blower cpu cools and bring that down a bit19:11
lazyPoweryeah man, looks like a solid layout from my perspective19:11
Budgie^Smoreoh and about another $80 for a 16 port managed switch19:11
Budgie^Smorejust haven't figured out the cost of the case19:12
shann_lazyPower, i think mount lab's with Juju, serve apps, but i think not use couple of public_ip:port, i think possibility to use endpoint proxy to redirect traffic on app19:12
Budgie^Smoreoh and I have designed it in a way that you can add more levels to increase the cluster size19:13
bdxBudgie^Smore: I like man ... super cool19:14
lazyPowerah we dont expose anything like that in juju. You can add users to the controller, and share the juju-gui url to your lab students. Then they can in turn manipulate and deploy/destroy applications in their model, using their individual apps public ip:port combination19:16
lazyPowershann_: would that satisfy the requirement?19:16
lazyPowerat least, we dont expsoe anything like that if i'm understanding correctly.19:17
lazyPowerallow me to be clear on that. i'm still a little confused as to what you're asking for19:17
shann_i found this, https://github.com/vtolstov/charm-haproxy, if i understood, it's define single endpoint.19:18
lazyPowershann_: ah yes, you can certainly use haproxy to reverse proxy into applications in a model19:18
shann_yes infact, but this configuration it's correctly usage with juju, or juju has own method implemented ?19:20
bdxshann_: possibly `juju deploy haproxy` is what you want19:22
shann_example haproxy can serve several apps relations ? ex haproxy => first_app, haproxy => second_app, ...., i try to understand archi before mount demo on my computer.19:23
shann_but i think better is test ^^19:24
Budgie^Smorebac the idea would be to do the design work and then try and crowd source a few purchases to bring the individual cost down19:27
bdxshann_: yeah it can do that ... but you have to write the code for the other side of the relation that connects to haproxy to correctly pass the information needed to haproxy charm to make it do what you want19:42
bdxshann_: so you can connect haproxy to a single application and it will automatically do what you want19:43
bdxreverse proxy to that app19:43
bdxbut it doesn't know what to do when you connect subsequent applications19:43
shann_bdx, yes infact seem work out of box for haproxy linked app. It's possibly wrong way for my reflexion.19:44
shann_i try to execute juju bootstrap localhost controller, but blocking with "Waiting for address" :(20:31
bdxshann_: do you have lxd installed and working? (does `lxc launch ubuntu:16.04 u1` work for you?)20:32
shann_bdx, yes u1 launched but not have ip :(20:34
shann_lxd installed, bridge lxdbr0 seem up, with ip defined with dpkg-reconfigure -p medium lxd20:34
shann_no missing deps (bridge-utils, dhcp,.. ?)20:35
bdxshann_: `lxc delete u1 --force`20:35
bdxthen20:36
bdxsudo lxd init20:36
bdxyou might need to remove the image too20:36
bdxlxc image delete juju/xenial/amd6420:37
bdxshann_: sudo service lxd-bridge restart; sudo service lxd restart20:39
shann_yes infact, i remove image, also i restart lxd.socket, seem "lxd init" warning them still activate20:39
shann_wait 5/10mns with my little connexion20:39
shann_400ko/s :'(20:40
shann_bdx, ok image download, but continue to blocked at "waiting for address :("20:47
bdxhmmm20:48
bdxshann_: will you paste the output of this command `cat /etc/default/lxd-bridge | pastebinit`20:49
shann_humm, in log i have "invalid pid for SIGHLD. Received pid yyy expected pid xxx" (for juju-<random_id>à20:49
shann_humm, in log i have "invalid pid for SIGHLD. Received pid yyy expected pid xxx" (for juju-<random_id>à20:49
shann_paste.ubuntu.com/2456301520:50
bdxshann_: looks good to me20:52
bdxhmmm20:52
shann_bdx, lxdbr0 bridge need matched with nic ?21:09
shann_brctl show, lxdbr0 interface <nothin>21:09
shann_i have enp0s... and wlp0s3 for wireless21:10
bdxshann_: it shouldn't be bridged to any interface I don't thing22:26
bdxcan anyone give me a quick run down on deploying windows with juju on aws?22:27
bdxis that a thing?22:27
Budgie^SmorelazyPower think you will appreciate the color now ;)23:02

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