Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel, 2016-02-10

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:41 wiza joined #fuel
00:56 xarses joined #fuel
01:08 ron___ joined #fuel
01:29 krobzaur_ joined #fuel
01:37 dslevin1 joined #fuel
01:44 xarses joined #fuel
01:57 Sparyx joined #fuel
02:41 wiza joined #fuel
04:25 claflico joined #fuel
05:13 javeriak joined #fuel
05:19 javeriak joined #fuel
06:46 javeriak joined #fuel
06:49 javeriak joined #fuel
06:51 javeriak_ joined #fuel
07:30 javeriak joined #fuel
07:55 magicboiz joined #fuel
07:58 e0ne joined #fuel
08:00 tzn joined #fuel
08:04 dslevin joined #fuel
08:20 fzhadaev joined #fuel
08:25 samuelBartel joined #fuel
08:27 zubchick joined #fuel
08:36 Philipp__ joined #fuel
08:40 neilus joined #fuel
08:46 zubchick joined #fuel
08:50 elopez joined #fuel
09:19 javeriak joined #fuel
09:45 tzn joined #fuel
10:02 javeriak joined #fuel
10:07 e0ne joined #fuel
10:25 javeriak_ joined #fuel
11:07 dklepikov joined #fuel
11:10 javeriak joined #fuel
11:25 akscram joined #fuel
11:45 az` joined #fuel
11:55 krypto joined #fuel
12:01 neilus1 joined #fuel
12:05 tzn joined #fuel
12:08 javeriak joined #fuel
12:09 krypto joined #fuel
12:17 Sparyx joined #fuel
12:19 krypto in network nodes i see
12:19 krypto "neutron-plugin-openvswitch-agent stop/waiting" but neutron agent-list shows it :) ,no interruption to services also,how is this achieved
12:32 tzn joined #fuel
12:35 krypto okk i think its because in network node its started through crm and in compute its through init
12:36 tzn joined #fuel
12:42 zubchick joined #fuel
12:53 zubchick joined #fuel
12:57 neilus joined #fuel
12:59 neilus joined #fuel
13:07 tzn joined #fuel
13:07 tzn joined #fuel
13:08 krypto joined #fuel
13:09 e0ne joined #fuel
13:17 az` joined #fuel
13:18 Sparyx joined #fuel
13:25 e0ne joined #fuel
13:34 krypto joined #fuel
13:38 xek joined #fuel
13:40 javeriak joined #fuel
13:42 dslevin left #fuel
13:55 javeriak_ joined #fuel
14:07 neilus1 joined #fuel
14:29 krobzaur_ joined #fuel
14:30 neilus joined #fuel
14:49 zubchick joined #fuel
15:13 claflico joined #fuel
15:22 xarses joined #fuel
15:31 Verilium As I understand it, Fuel doesn't currently support in-place environment upgrades, from one major release to another (Kilo -> Liberty).  I think it's a planned feature though?
15:32 mwhahaha yea we're working on it, not sure the time frame. ogelbukh might be able to provide more information
15:32 Verilium Ah, just found the blueprint page about it.  https://blueprints.launchpad.net/fuel/+spec/upgrade-major-openstack-environment
15:32 * Verilium nods.
15:34 blahRus joined #fuel
15:35 e0ne joined #fuel
15:52 dancn Verilium: you may have a look to <https://www.openstack.org/summit/austin-2016/vote-for-speakers/presentation/8159> I just have seen it today...
15:53 Sparyx joined #fuel
16:05 Verilium dancn:  Oh wow, nice, thanks for the link!
16:21 [HeOS] joined #fuel
16:43 zubchick joined #fuel
16:53 severion joined #fuel
17:09 zubchick joined #fuel
17:10 Verilium Booo, my boss won't send me to Austin. :P
17:21 thumpba joined #fuel
17:24 skamithi joined #fuel
17:31 skamithi hey folks. new to using Fuel. just wanted to share that I built virtual env using vagrant-libvirt to study it more. http://linuxsimba.com/mirantis.html. btw - really cool GUI. topology is pretty basic but hope to add some virtual switches (cumulus vx)  and document that experience in a future blog post.
17:40 fedexo joined #fuel
17:41 fedexo_ joined #fuel
17:42 Sparyx joined #fuel
17:42 mwhahaha skamithi: cool, we also have some scripts for building and managing environments on libvirt - https://docs.fuel-infra.org/fuel-dev/devops.html
17:43 skamithi cool. i'll check those out. i only found the virtualbox scripts
17:44 mwhahaha we use the devops scripts in development and testing
17:46 elopez joined #fuel
17:47 mugleeta joined #fuel
17:51 krobzaur joined #fuel
18:04 javeriak joined #fuel
18:19 xarses skamithi: we use the fuel-devops scripts to control libvirt, but it's intended to work with qemu. Their might be bugs with libvirt-virtualbox but it would be awesome to get support there so we can maybe simplify some of the virtual box stuff
18:20 e0ne joined #fuel
18:25 slo_ joined #fuel
18:26 tzn joined #fuel
18:26 skamithi xarses: if i have time i'll try and do it in vagrant-virtualbox.  i'm not a virtualbox expert so I struggle with vagrant-virtualbox. already got a hashicorp Packer script for a fuel vagrant box. today only tested for building a qemu vagrant box, but don't see why it cannot work with virtualbox as well. Just lightly modified the ks.cfg script from the iso to support vagrant.
18:27 tzn joined #fuel
18:28 slo_ Quick question. Does MOS 7 allow to install compute and controller service on the same node. Let's say i have 5 node cluster. I want all nodes having compute and storage services, but just three of them would be controller nodes? Is it possible?
18:29 _tzn joined #fuel
18:30 slo_ Also. Is it somehow possible to install MOS 7 on two nodes (both having controller, compute, storage services)? Knowing that you are less prone for failures...since just two nodes would be in HA?
18:37 mwhahaha no
18:38 mwhahaha controllers need to be in an odd numbering for proper HA and you can't install controller & compute on the same node
18:40 slo_ mwhahaha: also no one tried to patch to do so till now? :)
18:40 slo_ i know there was a chance to put all on one note, but i guess just with simple, flat network and no ceph, i guess lvm or swift used as backend?
18:41 slo_ i am not sure if that was mos 7.0, probably 6.1
18:42 slo_ is that even considerable within telco environments, because i guess they want to have one platform for all kind of installation....knowing the limitations for HA and reliability...
18:47 mwhahaha well it's not a good idea for security reasons
18:47 slo_ joined #fuel
18:47 mwhahaha since compute is where tenant untrusted vms are going to be run, you shouldn't allow it on your controllers which house your auth infrastructure
18:50 slo_ ok
18:51 slo_ but let's say that all VMs are strictly under the control of the operator? telco app
18:51 mwhahaha yea still not a good idea :D
18:51 slo_ would that make sense?
18:51 slo_ aha
18:51 mwhahaha as an operator, they should trust the devs as much as 3rd party customers. ie, nope
18:52 slo_ haha :)
18:52 slo_ nice one
18:55 slo_ ok thx for now. i need to reconsider my idea :)
19:10 thumpba joined #fuel
19:25 slo_ joined #fuel
19:54 xarses slo_: you can have as few as 1 controllers, and add more later
19:54 xarses but no, we don't offer combined controller and compute node
19:55 xarses there is a reduced footprint feature that could be used to create vm's on libvirt-qemu-kvm nodes and then I think you can also then allocate those nodes as compute nodes
20:00 Reepicheep joined #fuel
20:16 slo_ @xarses: ok. good to know. is there any blog about that doing somewhone that?
20:17 slo_ but in case i choose one controller and one compute node, let's say two node cluster, than ceph is anyway not an option, so LVM is used (possible with drbd)
20:18 xarses ceph is an option
20:18 xarses you can run the osd on the controller and compute node
20:18 xarses you just have to assign them seperatly
20:18 xarses and you need to set the replica count to 2 from 3
20:19 xarses slo_:  http://www.yet.org/2015/10/mos7-reducedfootprint/
20:20 zubchick joined #fuel
20:21 slo_ thank you for the link!
20:26 zubchick joined #fuel
20:28 slo_ Will this feature also be available in MOS 8?
20:31 zubchick joined #fuel
20:39 angdraug joined #fuel
20:42 kozhukalov joined #fuel
20:42 HeOS joined #fuel
20:44 xarses yes
20:46 tzn joined #fuel
20:51 tzn joined #fuel
20:53 e0ne joined #fuel
21:01 kozhukal` joined #fuel
21:15 HeOS joined #fuel
21:18 kozhukalov joined #fuel
21:19 slo_ xarses: one more thing, so for Ceph one MON is enough? Isn't Paxos consensous algorithm a little less decisive than or is it ok?
21:20 slo_ since mirantis MOS deployment puts Ceph's MONs just on controller nodes, which is in my case, 1
21:26 xarses slo_: it will work you just won't have any real failure protection untill you have an odd number of controllers above 2
21:27 slo_ ok, understood.
21:27 slo_ so no rebalancing od data in case of disk fialed , happens
21:27 xarses the story is similar for other components on the controller besides ceph-mon
21:27 slo_ but writing two times is possible
21:27 slo_ ok
21:27 slo_ also with mysql-galera i guess
21:28 xarses the mon's don't participate in a rebalance other than re-calculating placement groups
21:29 slo_ let' say i have 4 disks in each node. in case of one disks fails, would the replica being made by primary OSD or not?
21:29 xarses the number of mon's isn't relevant to data protection in the OSD's
21:30 xarses that's the replica factor, number of OSD nodes, and CRUSH map
21:30 xarses so with a replica factor of 2 the default crush map will pace a copy on two hosts
21:30 slo_ ok
21:30 xarses which makes at least two ceph-osd roles required
21:31 slo_ cool
21:32 xarses the default replica count is 3, which makes fuel require you to allocate at least 3 OSD hosts
21:33 xarses you can have more OSD's on each host than that, but the default CRUSH map will not allocate replicas from the same placement group to the same host
21:35 slo_ that's ok. since there is no really HA from the HW perspective (2 nodes), there is not to be expected better availability of the data , so exceptable :)
21:35 xarses having data durability is still desirable
21:35 xarses if you dont care, set your replica count to 1
21:36 slo_ yes, but if one want's that kind of configuration, it have a solution, but with boundaries that are known :)
21:40 mrasskaz- joined #fuel
21:41 v1k0d3n_ joined #fuel
21:41 charz_ joined #fuel
21:41 neouf_ joined #fuel
21:44 obcecado_ joined #fuel
21:44 damjanek_ joined #fuel
21:47 HeOS joined #fuel
21:49 tlbr joined #fuel
21:54 t_dmitry_ joined #fuel
21:56 as0bu joined #fuel
21:56 mkwiek joined #fuel
21:57 justif joined #fuel
21:57 mmalchuk joined #fuel
21:57 skath_ joined #fuel
22:18 Reepicheep_ joined #fuel
22:23 Sesso joined #fuel
22:31 babayarba joined #fuel
22:31 babayarba hey guys
22:32 babayarba i _think_ i may have hit a bug on fuel-7. even though network test is green the deployment fails with a timeout to hit mysql for the cvs file.
22:33 babayarba sorry, for the csv
22:35 babayarba im using the Neutron with VLAN segmentation option
22:35 kozhukal` joined #fuel
22:41 mwhahaha does your public network have a pingable gateway?
22:52 babayarba mwhahaha: checking
22:53 babayarba should it?
22:53 mwhahaha yes
22:55 * babayarba nods
22:55 mwhahaha the controllers use it as a health check so if it doesn't what happens is the public vips get turned off which can lead to timeout issues
22:55 mwhahaha you can check pcs status on the controller(s) to see if the vips are up
22:57 babayarba vips are listed as "started"
22:58 babayarba but pcsd status - all three ips are listed as offline.
23:00 babayarba ah. it actually breaks earlier, trying to restart haproxy ten times.
23:02 babayarba never you mind. ssl config issue.
23:13 babayarba there's no validation for the custom cert files in the ssl configuration area, eh?
23:34 Sparyx joined #fuel
23:35 thansen joined #fuel
23:36 mwhahaha validation how so?
23:38 babayarba if i were to upload, say, an empty file, or a malformed file would fuel barf?
23:39 mwhahaha not sure, for some reason i doubt we're checking it
23:43 babayarba yeah so it seems that this gets checked really late in the game and if that file is not formed correctly haproxy cannot restart and the deployment fails in a really non obvious way
23:45 mwhahaha yea that should totally be a bug
23:54 babayarba granted there really is no validation
23:54 babayarba i could be triggering this behavior because i am hitting deploy from another desktop.
23:55 babayarba which would also be a bug but a far less obvious usecase

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary