Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel-dev, 2015-07-23

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

All times shown according to UTC.

Time Nick Message
00:07 xarses joined #fuel-dev
01:34 xarses joined #fuel-dev
01:41 xarses_ joined #fuel-dev
02:08 tzn joined #fuel-dev
02:26 gongysh joined #fuel-dev
03:22 neophy joined #fuel-dev
03:51 andreww joined #fuel-dev
04:01 tzn joined #fuel-dev
05:08 nurla joined #fuel-dev
05:12 monester joined #fuel-dev
05:28 amnk joined #fuel-dev
05:48 stamak joined #fuel-dev
05:51 amnk joined #fuel-dev
06:00 ddepaoli joined #fuel-dev
06:15 Miouge joined #fuel-dev
06:17 alex_bh joined #fuel-dev
06:21 mkwiek joined #fuel-dev
06:32 aleksandr_null joined #fuel-dev
06:32 ub joined #fuel-dev
06:54 salmon_ joined #fuel-dev
07:00 tzn joined #fuel-dev
07:00 tzn joined #fuel-dev
07:06 fuel-slackbot joined #fuel-dev
07:07 monester joined #fuel-dev
07:12 dancn joined #fuel-dev
07:13 mkwiek joined #fuel-dev
07:14 neophy joined #fuel-dev
07:14 e0ne joined #fuel-dev
07:32 DanyC sgolovatiuk: the GW is the same from both and i'm sure i have only 1 DHCP server
07:35 DanyC sgolovatiuk: what i might try now is to break the shared network and end up with 1 network per tenant pointing to the same provider network vlan, controlling the dhcp pool range to not clash
07:42 dkusidlo joined #fuel-dev
07:49 luqas joined #fuel-dev
07:56 subscope joined #fuel-dev
07:59 kozhukalov guys, let's merge this https://review.openstack.org/#/c/203499
08:00 hyperbaba joined #fuel-dev
08:01 nurla joined #fuel-dev
08:01 evgenyl kozhukalov: done
08:14 amnk joined #fuel-dev
08:21 kozhukalov evgenyl, thanx
08:25 subscope joined #fuel-dev
08:26 dkusidlo joined #fuel-dev
08:27 nurla joined #fuel-dev
08:44 kozhukalov guys, what do you think of substituting fuelmenu with just thoroughly commmented text file + vi ?
08:54 evgenyl kozhukalov: I like the idea, but it's easy to make typo, and you will have to reinstall your entire master.
08:55 ogelbukh this is awesome idea
08:56 ogelbukh and it's not that difficult to reinstall the master :)
08:56 ogelbukh I personally 100% prefer vim UX to fuelmenu :)
08:57 kozhukalov evgenyl, we can check syntax, can't we? it's much easier. is you ask me, the main pro of getting rid or fuelmenu is it definitely makes it easy to extend this text file, and it is easy for a user to edit it instead of going through semigraphycal menu.
08:58 kozhukalov s/is you ask me/if you ask me/
08:59 evgenyl kozhukalov: yeah, we can check syntax, and it's much more scripting friendly.
08:59 kozhukalov ok, then we need to raise this in openstack-dev ML
09:00 evgenyl kozhukalov: but as far as I remember there was some advanced validation which used network checker or something like that, and it looks overcomplicated.
09:00 evgenyl kozhukalov: yeah, lets discuss it in ML
09:01 diurchenko joined #fuel-dev
09:27 HeOS joined #fuel-dev
09:38 stamak joined #fuel-dev
09:46 dkusidlo joined #fuel-dev
09:56 evgenyl kozhukalov: I have a question regarding to our provisioning, Mike had a suggestion to think a bit if it makes sense to do partitioning and provisioning itself in two steps (two different tasks), so for example if you need only partitioning you can it without provisioning and vice versa.
09:57 kozhukalov evgenyl, no problem at all, fuel agent has a set of do_* methods which do those separate tasks
09:59 kozhukalov for example  https://github.com/stackforge/fuel-agent/blob/master/fuel_agent/manager.py#L113 makes partitions and provisioning itself calls other do_methods https://github.com/stackforge/fuel-agent/blob/master/fuel_agent/manager.py#L493
09:59 evgenyl kozhukalov: so basically I can run provisioning with volume as a parameter in order install OS, or how does it work?
10:00 evgenyl kozhukalov: ok, not only volume, there are many other parameters probably.
10:01 kozhukalov provisioning itself assume the whole flow including decommission, partitioning, downloading and putting images on hard drive
10:01 kozhukalov but if you already have necessary partitions, you can just call do_copyimage https://github.com/stackforge/fuel-agent/blob/master/fuel_agent/manager.py#L310
10:02 kozhukalov yes, all manager methods are data driven, and you can call them separately passing some input data
10:03 kozhukalov we probably need to modify fuel agent a little bit to address this context, but all the changes are gonna be minor
10:03 prmtl joined #fuel-dev
10:05 evgenyl kozhukalov: that is cool, I think volume manager should be able to send tasks to do only partitioning, and in this case, our provisioning will just consist of several tasks from different extensions.
10:09 romcheg joined #fuel-dev
10:28 kozhukalov evgenyl, yes, that would be cool
10:29 kozhukalov but need keep in mind how stable such an approach would be
10:31 dkusidlo joined #fuel-dev
10:32 evgenyl kozhukalov: what do you mean?
10:37 kozhukalov i mean if it is safe enough to separate those tasks so they are split in time. for example when the provisioning task is monolithic and run as a single process, it is unlikely that between partitioning and copying image partitions are changed. so can be sure that the partition where we are going to put an image exists. when these tasks are separate, we can not be sure 100% that input data for partitioning are consistent with copyimag
10:37 kozhukalov e input data and we can not be sure that partitions are not changed between these two separate tasks.
10:49 xenolog13 joined #fuel-dev
10:52 dims_ joined #fuel-dev
10:57 sanek joined #fuel-dev
11:08 tzn joined #fuel-dev
11:21 aglarendil joined #fuel-dev
11:27 stamak joined #fuel-dev
11:33 amnk joined #fuel-dev
11:36 nurla joined #fuel-dev
11:37 romcheg joined #fuel-dev
11:41 nurla joined #fuel-dev
11:42 xarses_ joined #fuel-dev
11:44 tlbr joined #fuel-dev
11:56 jaypipes joined #fuel-dev
12:52 alex_bh joined #fuel-dev
12:55 tzn joined #fuel-dev
12:59 samuelBartel joined #fuel-dev
13:09 mattymo_ joined #fuel-dev
13:17 ogelbukh guys, small question
13:17 ogelbukh where does the 'extra' attribute of disks come from?
13:17 ogelbukh can't find it in fuel-agent or nailgun
13:17 ogelbukh ^^ kozhukalov Vladimir, could you tell?
13:28 yottatsa joined #fuel-dev
13:30 dkusidlo joined #fuel-dev
13:31 yottatsa joined #fuel-dev
13:31 ogelbukh ok, nevermind, found it
13:32 agordeev ogelbukh: it comes from udev properties. Some values may be filtered out.
13:34 agordeev ogelbukh: nailgun-agent. https://github.com/stackforge/fuel-nailgun-agent/blob/master/agent#L387
13:37 ogelbukh so another question, how this agent relates to fuel-agent?
13:39 agordeev ogelbukh: fuel-agent does provisioning and image building. nailgun-agent does discovery of node (eg: reporting node capabilities such as cpu, ram, disk) to nailgun API
13:40 ogelbukh ok good
13:40 ogelbukh thank you
13:42 agordeev all data about the node is collected by nailgun-agent, after that it will be put in nailgun. Nailgun will generate disk partitioning scheme and other thing from this data and pass it to fuel-agent to let it provision a node.
13:42 ogelbukh understood
13:44 dims_ joined #fuel-dev
13:44 agordeev ogelbukh: https://docs.mirantis.com/openstack/fuel/fuel-6.1/reference-architecture.html#fuel-agent-arch 'Operating system provisioning' describes that
13:45 dims_ joined #fuel-dev
13:56 jaypipes joined #fuel-dev
13:59 yottatsa joined #fuel-dev
14:03 ogelbukh good
14:04 ogelbukh please look into this issue then https://bugs.launchpad.net/fuel/+bug/1477604
14:04 ogelbukh it seems to be a root cause for many bugs associated with refactoring of volume manager
14:05 evgenyl We should rename fuel-agent into something more specific, it confuses when there are two agents.
14:05 ogelbukh evgenyl: +1
14:08 claflico joined #fuel-dev
14:20 rbrooker joined #fuel-dev
14:35 agordeev ogelbukh: fuel-agent pays attention to 'extra'. https://github.com/stackforge/fuel-agent/blob/master/fuel_agent/drivers/nailgun.py#L53-L66
14:36 ogelbukh that's right, but volume manager doesn't
14:36 romcheg1 joined #fuel-dev
14:36 ogelbukh and UI and API work with volume manager
14:36 ogelbukh I ended up in a situation recently when I can't upload back what I've downloaded from API
14:37 ogelbukh because disk ids changed
14:38 evgenyl ogelbukh: so the problem is after node is provisioned, disks in vm api get mixed, correct?
14:38 ogelbukh yes, that's correct
14:39 ogelbukh I'll send a small change to try to fix that
14:39 evgenyl ogelbukh: I think nobody tried to do it before, I mean to configure disks, when node is already provisioned.
14:39 ogelbukh probably, ues
14:39 ogelbukh *yes
14:39 ogelbukh but it's not only configure
14:40 ogelbukh imagine if I want to delete the node and install it in another env with the same disk config
14:41 ogelbukh in theory, I should be able to just get in from API and send it back once the node is moved
14:41 evgenyl ogelbukh: oh, now I see what case causes the problem.
14:41 ogelbukh generally speaking, it's when I try to use Fuel as a source of truth for hw config :)
14:42 ogelbukh which could be not intended use, I admit :)
14:42 evgenyl :)
14:43 evgenyl ogelbukh: is there a way to identify volumes differently, as far as I know uniqueness can differ from distributive to distributive.
14:43 ogelbukh but seriously guys, it's a hw inventory, want it or not
14:44 ogelbukh well, at least in ubuntu /dev/disk/by-id seems to be more reliable way
14:44 ogelbukh and it's already in metadata actually
14:44 ogelbukh another way to fix it is using udev that is not affected by the bug in question
14:44 evgenyl kozhukalov: ^ could you please comment on that, as far as I know you researched this problem some time ago.
14:45 ogelbukh because it's clearly a bug that udev just ignores SCSI numbers reported by driver
14:45 agordeev ogelbukh: at least different kernel/udev versions could lead to this kind of disk inconsistency.
14:45 evgenyl ogelbukh: we can fix it in with ugly hack, which is stop updating information about disks after node was provisioned.
14:45 agordeev bootstrap image is still on centos6.5, but there's a bp
14:46 ogelbukh evgenyl: it is already implemented
14:46 agordeev https://blueprints.launchpad.net/fuel/+spec/fuel-bootstrap-on-ubuntu
14:46 ogelbukh but it doesn't get us closer to lifecycle management way of managing things ;)
14:46 evgenyl ogelbukh: but in this case you should be able to get non-mixed disks and recreate the node.
14:46 evgenyl ogelbukh: absolutely agree.
14:47 evgenyl ogelbukh: that is why I call it "ugly hack" :)
14:47 ogelbukh right
14:47 ogelbukh OK let's think a bit more about options here
14:47 ogelbukh first of all I'd like to understand what is affected and what's not
14:50 kozhukalov ogelbukh, let's have a meeting
14:50 kozhukalov it certainly will take less time
14:51 kozhukalov i am not even able to read all this stuff that you wrote -)
14:53 evgenyl :)
15:02 alex_bh joined #fuel-dev
15:25 rmoe joined #fuel-dev
15:38 rbrooker joined #fuel-dev
15:51 xarses_ joined #fuel-dev
15:52 tatyana joined #fuel-dev
15:53 alexz joined #fuel-dev
15:58 skath joined #fuel-dev
15:58 jaranovich joined #fuel-dev
15:59 kat_pimenova joined #fuel-dev
16:00 kpimenova joined #fuel-dev
16:00 jaranovich left #fuel-dev
16:01 warpc__ joined #fuel-dev
16:02 angdraug joined #fuel-dev
16:06 ogelbukh :)
16:07 justif joined #fuel-dev
16:12 [HeOS] joined #fuel-dev
16:30 dpyzhov_ joined #fuel-dev
16:43 amnk joined #fuel-dev
16:51 teran joined #fuel-dev
17:02 jobewan joined #fuel-dev
17:24 neophy joined #fuel-dev
17:33 teran joined #fuel-dev
17:35 e0ne joined #fuel-dev
17:44 romcheg joined #fuel-dev
17:45 dims joined #fuel-dev
17:56 angdraug joined #fuel-dev
17:56 romcheg1 joined #fuel-dev
17:58 romcheg joined #fuel-dev
18:11 amnk joined #fuel-dev
18:18 yottatsa joined #fuel-dev
18:41 Miouge joined #fuel-dev
18:42 e0ne joined #fuel-dev
18:55 bitblt joined #fuel-dev
19:01 amnk joined #fuel-dev
19:09 ub joined #fuel-dev
19:16 jaypipes joined #fuel-dev
19:20 tzn joined #fuel-dev
19:29 teran joined #fuel-dev
19:37 Miouge joined #fuel-dev
20:01 Miouge joined #fuel-dev
20:03 ub2 joined #fuel-dev
20:08 mihgen folks I just spoke to alex_didenko, https://review.openstack.org/#/c/204111/ is gonna be merged, and Alex will write email about it to openstack-dev
20:09 mihgen we've got +1th from keystone team
20:09 mihgen FYI in case anyone has any objections, I believe this is important patch - it's Apache for keystone
20:19 ub joined #fuel-dev
20:28 mwhahaha one comment
20:32 mwhahaha i'm ok now
20:38 ub joined #fuel-dev
20:50 e0ne joined #fuel-dev
21:06 ub2 joined #fuel-dev
21:12 e0ne joined #fuel-dev
21:25 HeOS joined #fuel-dev
21:36 HeOS joined #fuel-dev
21:48 tzn joined #fuel-dev
22:00 HeOS joined #fuel-dev
22:14 HeOS joined #fuel-dev
22:17 ub joined #fuel-dev
22:39 dims joined #fuel-dev
22:46 HeOS joined #fuel-dev
23:04 teran_ joined #fuel-dev
23:06 teran joined #fuel-dev
23:14 kaliya joined #fuel-dev
23:24 dims_ joined #fuel-dev
23:34 preilly joined #fuel-dev

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