| === Spads_ is now known as Spads | ||
| === frankban|afk is now known as frankban | ||
| kjackal | Good morning Juju world! | 07:29 |
|---|---|---|
| roadmr | hello folks! Hey, periodically I have to "compact" my juju1 db, particularly on node 0. For this I need to stop the juju-db service. Is there a way to compact the DB while keeping juju running? | 14:14 |
| rick_h | roadmr: sorry nothing new on that front. There's a ton of work into 2.2 for db size and maintenance but nothing that would compact while it's running. I think that's more a mongodb limitation that the db has to come offline to do the compacting. | 14:16 |
| roadmr | rick_h: ok, it's what I thought but felt it couldn't hurt to ask :) | 14:16 |
| === med_` is now known as med_ | ||
| rick_h | roadmr: definitely | 14:17 |
| roadmr | rick_h: thanks though! | 14:17 |
| lazyPower | rick_h: i wonder if thats not something we could tune for a maintenance window? | 14:18 |
| lazyPower | something like ship w/ a cron job that once a week takes it offline for 15 minutes to compact teh db | 14:18 |
| rick_h | lazyPower: well I think the goal is that you won't need compacting with all the work going into maintaining the db better | 14:19 |
| lazyPower | unless thats taking too many liberties.... in which case i would then think modeling the controller as a charm and giving it some actions would be ace | 14:19 |
| rick_h | lazyPower: but it'd be something that'd be interesting to see if we can get concrete numbers on how/when a user would need to. | 14:19 |
| rick_h | lazyPower: :) well that's the long term idea | 14:19 |
| lazyPower | aww yeee | 14:19 |
| lazyPower | <3 | 14:19 |
| lazyPower | rick_h: i redeployed my k8s cluster this weekend (in house) so i could really change teh topology. I'm happy to report there were no issues sticking flannel on the controller ndoe | 14:20 |
| lazyPower | i did some very dirty things... | 14:20 |
| lazyPower | stuff you would approve of, but it all worked | 14:20 |
| lazyPower | so all our effort in making 2.x more bulletproof has paid off | 14:20 |
| rick_h | lol | 14:20 |
| SimonKLB | anyone know how one would go about co-locating in amulet tests? since you can't add an application after you run setup() im not sure if there is any good way to find the correct machine id for placement | 14:32 |
| SimonKLB | libjuju to the rescue :D | 14:44 |
| lazyPower | SimonKLB: you read my mind :) | 15:46 |
| kim0 | Howdy channel .. I saw some discussion about cloud-integration (cloud-native) .. Can someone give me an idea what that is about .. Thanks! | 16:05 |
| lazyPower | kim0: with respect to kubernetes? | 16:57 |
| kim0 | yes | 16:58 |
| lazyPower | kim0: certainly. Cloud integration would bridge the gaps between CDK and what K8s natively supports. Today we model Kubernetes in a bare-metal first nature, where it behaves the same as it does on bare metal no matter where you deploy it. Many users have requested cloud-integration for supported things like Service type: load-balancer which will yield an ELB on AWS | 16:59 |
| lazyPower | auto provisioning of EBS storage as PV's | 16:59 |
| kim0 | Aha I see .. yeah that would surely be useful :) | 16:59 |
| lazyPower | things of that nature. So our goal there is to model it in a secure way that wont leak credentials, and allow the administrator to plug those cloud-integrations right into CDK for a seamless experience. Bare metal users are still tended to as they just dont enable the intgrations. | 16:59 |
| kim0 | is that coming any time soon :) | 17:00 |
| lazyPower | we have some pre-alpha quality work now, nothing public that I can point you to | 17:00 |
| lazyPower | but it is on the docket for this 17.10 cycle | 17:00 |
| === frankban is now known as frankban|afk | ||
| Budgie^Smore | lazyPower does it usually take this long to hear back about that developer access? | 17:42 |
| lazyPower | Budgie^Smore: no | 17:51 |
| lazyPower | you should have heard back already. I know that marco is currently at ODS | 17:51 |
| lazyPower | so that may be why, it got missed and he's busy giving talks. | 17:51 |
| lazyPower | Budgie^Smore: i'll poke him gently and see if we can expedite that | 17:51 |
| Budgie^Smore | lazyPower no problem was just curious... probably right con-season can be crazy | 17:52 |
| === rahworkx is now known as rahworks | ||
| Budgie^Smore | I really want to build a "cluster in a box" that could be "picked up" by devs / iot people | 18:15 |
| Budgie^Smore | http://www.asrockrack.com/general/productdetail.asp?Model=E3C226D2I looks a nice mini itx with ipmi support! | 19:04 |
| lazyPower | Budgie^Smore: thats a nice lookin box bruv | 19:20 |
| lazyPower | s/box/board | 19:20 |
| Budgie^Smore | yeah 2 NICs and 1 IPMI separate is really nice to find | 19:21 |
| Budgie^Smore | only real prob right now is to find a small yet large enough managed switch (12 port min) | 19:23 |
| lazyPower | that doesn't break the bank or require a unifi controller amirite? :D | 19:23 |
| Budgie^Smore | something like that... really wondering if I could "crowdsource" this idea maybe even build some sort of backplane that connects multiple boards with a switch | 19:25 |
| Budgie^Smore | I can't be the only one that would drop a couple of grand on a "cluster in a box" solution | 19:26 |
| lazyPower | Budgie^Smore: you're not, thats why we built the orange box | 19:44 |
| lazyPower | Budgie^Smore: a lot of vendors are showing up at conferences with ARM based alternatives like cubie boards or rpi clusters. | 19:45 |
| Budgie^Smore | yeah and discontinued it too :-/ although I would say that box was a little bit over powered for a small cluster | 19:45 |
| lazyPower | I myself have a stack of about six rpi 2's and 3's next to me on my desk that I fiddle around with periodically. | 19:45 |
| lazyPower | well, you wanted some beef to run the workloads + openstack | 19:45 |
| Budgie^Smore | oh yeah for conference, etc. it would be awesome but I am thinking more small form factor 5 node box | 19:46 |
| lazyPower | its less compelling if its just an openstack box, but if you can deploy stuff on top of it and demo the whole suite of jujucharms on sight without dealing with firewalls, access, et-al. its a handy tool for tech previews | 19:46 |
| lazyPower | I hear ya. Thats the entire reason i dubbed my pi cluster "The Clementine Box" | 19:46 |
| Budgie^Smore | 1 node maas + 4 nodes whatever the f you want ;-) | 19:46 |
| lazyPower | i wanted something to stand up both lxd and swarm on and demo the two container techs working in harmony | 19:47 |
| lazyPower | and they do :D | 19:47 |
| lazyPower | but provisioning is not as nice as PXE only landed on the rpi3 | 19:47 |
| lazyPower | and power management is still an issue | 19:47 |
| Budgie^Smore | yeah I still want an x86 platform | 19:47 |
| lazyPower | but the baseline concept is there and it was a great learning tool | 19:47 |
| lazyPower | i stll say grab some gigabyte brix boards | 19:47 |
| lazyPower | and hobby to your hearts content | 19:47 |
| lazyPower | they're pretty small and beefy enough to run some serious workloads on them, thats 3/4 of my @ home k8s cluster is gigabyte brix units. | 19:48 |
| Budgie^Smore | hence looking at that board... having 2 gbic nic and ipmi gives you everything you really need for both power management and network layouts | 19:48 |
| Budgie^Smore | pop in an i3 or i5 and you have a beefy little server cluster | 19:48 |
| lazyPower | yep yep | 19:55 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!