Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel-dev, 2014-05-30

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

All times shown according to UTC.

Time Nick Message
00:17 e0ne joined #fuel-dev
00:42 rmoe joined #fuel-dev
00:46 IlyaE joined #fuel-dev
00:48 e0ne joined #fuel-dev
02:14 fandi joined #fuel-dev
04:08 tnurlygayanov joined #fuel-dev
04:58 IlyaE joined #fuel-dev
05:47 dshulyak_ joined #fuel-dev
05:51 tnurlygayanov joined #fuel-dev
06:19 salmon_ joined #fuel-dev
06:30 sambork joined #fuel-dev
06:42 bogdando joined #fuel-dev
06:46 e0ne joined #fuel-dev
06:58 artem_panchenko joined #fuel-dev
07:02 e0ne joined #fuel-dev
07:13 e0ne joined #fuel-dev
07:17 e0ne_ joined #fuel-dev
07:27 IvanBerezovskiy joined #fuel-dev
08:21 teran joined #fuel-dev
08:35 e0ne joined #fuel-dev
08:42 e0ne joined #fuel-dev
08:42 brain461 joined #fuel-dev
09:02 teran joined #fuel-dev
09:11 brain461 joined #fuel-dev
09:31 sambork joined #fuel-dev
09:33 ikalnitsky joined #fuel-dev
09:33 ikalnitsky left #fuel-dev
09:33 ikalnitsky joined #fuel-dev
09:40 EeeBotu New bug #1324830 in fuel: "FUEL Web UI with NSX selected  does not close its window" [Undecided, New] https://bugs.launchpad.net/bugs/1324830
10:05 salmon_ mihgen: is there anyone who can design login page for  us?
10:08 vkramskikh salmon_: let me read the final version of blueprint
10:08 salmon_ vkramskikh: 1h :)
10:08 vkramskikh we ended up with the decision that we need to protect both UI and API, right?
10:09 salmon_ vkramskikh: yes, but we have to split it into smaller stages. I'm writing new version now and will share it
10:09 vkramskikh okay
10:10 vkramskikh i don't think login page will be that complex so it needs to be thoroughly designed :) anyway i think b-zone will help us :)
10:11 salmon_ vkramskikh: yeah, it should be easy, but it should look nice
10:23 b-zone i'm here.
10:23 tnurlygayanov Hi there!
10:24 tnurlygayanov I have strange pproblem with my envrinoement with 50 hardware servers
10:24 sambork joined #fuel-dev
10:25 tnurlygayanov When I remove my environment with 50 hardware nodes, part of my slave nodes don't come back to Fuel master node
10:25 tnurlygayanov so, after that I can see only part of my nodes: 33 nodes or 45 nodes.
10:26 tnurlygayanov and I should reboot my nodes, which are losted by Fuel master - after that I can see all slave nodes in Fuel
10:27 tnurlygayanov can anybody suggest what the root of this problem and how I can fix it? :) I can provide the access to my environment for debugging
10:30 teran_ joined #fuel-dev
10:30 teran_ joined #fuel-dev
10:40 teran joined #fuel-dev
10:49 EeeBotu New bug #1324859 in fuel: "Could not start Service[neutron-server]" [High, New] https://bugs.launchpad.net/bugs/1324859
10:49 EeeBotu New bug #1324850 in fuel: "FUEL deployment fails with NSX" [Undecided, New] https://bugs.launchpad.net/bugs/1324850
11:13 zinovik joined #fuel-dev
11:40 evg tnurlygayanov: hi, so your nodes can't pxe boot simultaneously? Could it be that something's changed in linux kernel in fuel5.0?
11:41 mattymo tnurlygayanov, concurrent bootstrap deploy is slow. Not something we can fix easily
11:42 tnurlygayanov evg, mattymo, yes, looks like it is problems with pxe/bootstrap
11:42 evg tnurlygayanov: Try tunning the kernel's network stack. See https://bugs.launchpad.net/fuel/+bug/1306705 (the end of the thread)
11:43 mattymo you need hardware to redeploy more than 20 at a time
11:44 tnurlygayanov so, it is cool baremetal servers with 16 cores...
11:45 tnurlygayanov evg, does it will help: sysctl -w net.core.rmem_max=26214400   ?
11:45 tnurlygayanov just one command of Fuel master node?
11:46 salmon_ vkramskikh: review https://review.openstack.org/#/c/96429/2 and here is nice looking version http://docs-draft.openstack.org/29/96429/2/check/gate-fuel-specs-docs/91779b1/doc/build/html/specs/5.1/access-control-master-node.html
11:47 salmon_ mihgen: dpyzhov mattymo we added stages there ^
11:49 mattymo you're going to install a 2nd service in a container? why not separate container like our containerized architecture?
11:49 * mattymo -1s
11:49 salmon_ mattymo: it may be new container no problem
11:50 salmon_ are there any guidelines aobut this?
11:50 mattymo yeah I can walk you through how to create a new container and test it
11:50 mattymo we've made it pretty simple
11:50 salmon_ mattymo: sure
11:51 mattymo you need a puppet manifest to deploy just keystone and a dockerfile similar to https://github.com/stackforge/fuel-main/blob/master/docker/nailgun/Dockerfile
11:51 mattymo and we will likely add new fields and new defaults to astute.yaml
11:52 mattymo all that changes between containers is ports exposed, maybe some directory creation, and what puppet file to apply
11:53 salmon_ mattymo: look really nice
11:53 salmon_ *looks
11:55 e0ne joined #fuel-dev
11:56 mattymo salmon_, 3 points I commented on - you shouldn't find them too brutal
11:58 evg tnurlygayanov: really, i have tried, just point to possible issue. Yes, on master.
12:00 evg tnurlygayanov: to be sure, of couse, you should tcpdump the problem.
12:04 tnurlygayanov evg, ok, thank you!
12:06 evg tnurlygayanov: opps sorry, i mean "really, i have _not_ tried"
12:13 salmon_ mattymo: I added replies
12:14 mattymo salmon_, you didn't run git review. still says 2nd patch set only
12:14 salmon_ mattymo: I added comments only
12:14 mattymo I don't see these
12:15 salmon_ ah
12:15 salmon_ mattymo: now
12:15 e0ne_ joined #fuel-dev
12:15 mattymo it's not so obvious
12:15 salmon_ yeah
12:21 vkramskikh salmon_: do you want auth to be always enabled? even by default right after master node installation
12:22 salmon_ vkramskikh: yes, wieth default credentials (admin/admin)
12:22 mattymo how do you disable it temporarily then enable it later?
12:22 mattymo that's a tough solution
12:23 vkramskikh there could be a flag in nailgun db which would determine auth state - disabled or not
12:23 mattymo a null auth driver, then switch to keystone
12:23 mattymo got it
12:25 salmon_ hmm, but why? What is a use case?
12:27 vkramskikh someone could want it, or for testing purposes, i don't know :) UI can support both cases. always enabled auth is easier to implement, let's implement it first
12:28 salmon_ vkramskikh: I will add this to stage III
12:29 mattymo I actually agree with vkramskikh for the sake of staged deployment
12:29 mattymo what if we do it in pieces and find a critical bug along the way? we could just temporarily switch the build default to disabled auth.. the flag may be handy during merge phases
12:30 mattymo but I'm ok with a permanent on or off flag at beginning of deploy
12:30 salmon_ mattymo: good point
12:30 mattymo salmon_, we did one for docker
12:30 salmon_ ok Stage II :)
12:30 mattymo at enormous expense, mind you :)
12:33 salmon_ mattymo: so I can run fuel without docker?
12:34 mattymo if you make iso PRODUCTION=prod, I believe yes.... I'm checking fuel-main
12:35 mattymo actually it looks like that was pulled from what was merged
12:35 mattymo legacy is gone :)
12:35 vkramskikh i didn't use keystone, how would the login process look like for stages I and II?
12:35 vkramskikh for stage 1, we should just pass basic auth headers for every request as far as i understand
12:36 vkramskikh what about stage 2, how do we get the token?
12:36 salmon_ vkramskikh: no, for stage I just login. No headers modification. Api will not be protected in stage I
12:36 vkramskikh oh, ok
12:37 salmon_ vkramskikh: as simple, as possible
12:37 salmon_ api will be protected in stage II
12:37 vkramskikh then if we use basic auth just for HUI, there won't be login and logout buttons/pages/anything
12:38 vkramskikh as "loginness" will be determined by browser
12:38 salmon_ vkramskikh: yeah, but: 1. it looks ugly, 2. we still need to something else for more advanced stuff
12:38 vkramskikh and login/password will be entered in a separate browser dialog
12:38 salmon_ *to do
12:39 salmon_ vkramskikh: stage I is only to be sure that we have something working for 5.1
12:39 salmon_ I think we should be able to do stage II, but just in case...
12:40 tnurlygayanov hm... probably someone can help me with another issue on Openstack Havanna 2013.2.2
12:42 vkramskikh i don't really understand how auth would work for stage 1
12:42 vkramskikh do you want to implement it on nginx level?
12:42 vkramskikh how will you integrate nginx with keystone?
12:42 tnurlygayanov in neutron logs I can see the following errors: http://paste2.org/1UVg6MpW
12:42 e0ne joined #fuel-dev
12:44 piontec joined #fuel-dev
12:44 salmon_ vkramskikh: no. in Stage 1 you will see login page. You will authorize using keystone. Somewhere in js app token will be stored but not used. If token is missing you will show login page again.
12:44 salmon_ Adding cookies here may be good idea
12:44 vkramskikh hmm
12:44 piontec hi everyone
12:45 vkramskikh it seems by "basic auth" you don't mean "basic http auth"?
12:45 piontec i have a repeatable bug/problem with fuel 5.0 release
12:45 vkramskikh http://en.wikipedia.org/wiki/Basic_access_authentication
12:45 piontec every time i create a new env, I get a ""Your OpenStack environment has been created, but configuration failed. You can configure it manually." message
12:46 piontec The problem is, that there are missing networking setting
12:46 piontec i can't choose vlan range for neutron with vlan, there is no "test network" panel
12:47 piontec and so on
12:47 salmon_ vkramskikh: I know how it works. Stage I is just something similar
12:47 piontec i could reproduce the problem on VirtualBox too, but couldn't find the direct cause; it seems to depend on the access to the internet/public net
12:48 piontec but i might be wrong here
12:48 piontec does anybody now a solution to this problem?
12:48 piontec i've commented and submitted a dump on launchpad: https://bugs.launchpad.net/fuel/+bug/1305154
12:48 piontec *know
12:48 vkramskikh salmon_: in this case i suggest to eliminate the word "basic" from stage 1 because it confuses a lot. i thought you are going to implement basic http auth in stage 1 and then replace it with keystone token in the header on stage 2
12:50 salmon_ vkramskikh: ah, ok. I will remove it. My point here is that implementing basic http auth is dead end, so we should implement something what at the beginning is similar but can be extended .
12:50 EeeBotu New bug #1324914 in fuel: "Need to set production-oriented configuration parameters for Neutron" [High, Confirmed] https://bugs.launchpad.net/bugs/1324914
12:53 vkramskikh salmon_: will auth actually work on stage 1? if API is still unprotected. what would be the logic? show login page if there is no keystone token in cookies?
12:55 salmon_ vkramskikh: no, in Stage I it's only javascript level protection. For full protection Stage II is required
12:56 vkramskikh ok, is the logic I described correct? do you want to implement it somehow like that?
12:57 salmon_ vkramskikh: yes, for stage I it's correct
12:57 salmon_ vkramskikh: let's call it kids protection level :)
13:03 vkramskikh piontec: that really strange, you are not the first person who faced this bug, but i wasn't able to reproduce it :( could you please open the UI in chrome, press F12 to open chrome dev tools, switch to "network" tab in dev tools and make a screenshot of the browser right after the moment you get this error in the cluster creation wizard?
13:05 piontec vkramskikh: sure, in a minute
13:05 teran joined #fuel-dev
13:06 piontec vkramskikh: for me it's very strange i'm the first: I can't make the bug go away on hardware, and on VBox it was also painfull, but suddenly disapppeared after network reconfiguration on host
13:06 teran_ joined #fuel-dev
13:10 e0ne joined #fuel-dev
13:11 piontec vkramskikh: screenshot is on comments on launchpad
13:11 piontec vkramskikh: let me know if there's anything i can do to walk around that bug and continue deployment
13:13 vkramskikh piontec: could you please also open http://10.20.0.2:8000/api/clusters/6/attributes in browser and attach output to the bug
13:14 vkramskikh piontec: actually you can continue deployment. this error means that the UI wasn't able to modify the settings of the cluster according to your input, you can go to settings tab and make the same changes manually
13:14 vkramskikh but i'm really curious about the reason
13:16 piontec vkramskikh: but there's nothing i can do waith netwroking
13:16 piontec no neutron config options, no "verfication toolbox"
13:16 piontec i can't configure anything
13:16 piontec network related
13:17 piontec the header just says "neutron with undefined segmentation", even though i choosed VLAN (and it's still the same for GRE)
13:17 piontec so there is no way i can continue
13:17 vkramskikh ok, got it
13:17 piontec i tested the CLI tool - exactly the same situation
13:18 vkramskikh please attach output of 10.20.0.2:8000/api/clusters/6/attributes to the bug
13:18 piontec sure, one sec
13:21 piontec vkramskikh: done, check launchpad
13:21 vkramskikh piontec: thanks
13:26 vkramskikh piontec: could you please also attach output of /api/clusters/6/network_configuration/nova_network and /api/clusters/6 and a screenshot of network tab?
13:27 vkramskikh the only strange thing i see in the output you attached is an abscense of vlan_splinters section..
13:37 e0ne joined #fuel-dev
13:40 e0ne_ joined #fuel-dev
13:49 piontec vkramskikh: updated on launchpad
13:49 salmon_ mattymo: did you see my last comment?
13:50 mattymo salmon_, not yet
13:50 EeeBotu New bug #1324927 in fuel: "qemu vulnerabilities (CVE-2014-0150, CVE-2014-2894)" [Undecided, In Progress] https://bugs.launchpad.net/bugs/1324927
13:50 mattymo ok salmon_ you tell me how you propose to ask a user in fuelmenu for a password and inject that user/pass combination into your keystone container?
13:50 vkramskikh piontec: oh sorry, output of /api/clusters/6/network_configuration/neutron too please
13:51 mattymo or do you use a fixed password that is just as insecure as we currently have?
13:53 salmon_ mattymo: with command `keystone user-update --password 03c84b51574841ba9a0d8db7882ac645 admin`
13:54 salmon_ mattymo: and service token can be generated
13:57 piontec vkramskikh: done
13:58 vkramskikh piontec: strange.. segmentation_type presents here, bug shown as "undefined" on the UI
13:59 vkramskikh *but
13:59 IlyaE joined #fuel-dev
14:02 mattymo salmon_, so a hardcoded passphrase?
14:03 mattymo how does nailgun learn what it is?
14:03 mattymo or how does a user?:)
14:03 mattymo especially if they're separated into different containers
14:03 vkramskikh piontec: the only suggestion i have is to try to clean browser cache and try to create a cluster in private window of your browser
14:05 vkramskikh if you used 4.1 before on the same address, the old UI might be cached
14:06 salmon_ mattymo: default password for user admin will be 'admin'
14:07 vkramskikh piontec: so clean cache, open a new private window and try to create a new cluster
14:08 salmon_ mattymo: do you know how keystone work?
14:12 mattymo I know we set a user-specified account with admin/admin by default
14:12 mattymo and we pass it via puppet in astute.yaml
14:12 mattymo and we create this account exactly this way
14:12 mattymo I'm trying to indicate we should set up keystone in the same manner we set up keystone in openstack
14:13 salmon_ no, no astute.yaml
14:13 salmon_ no in astute
14:13 mattymo go look at /etc/astute.yaml on a deployed node at the "access" hash
14:13 salmon_ in astute, only service token
14:13 salmon_ grr, no. nothing in astute.yaml :)
14:14 salmon_ astute.yaml is on nodes
14:14 piontec vkramskikh: oh sh***, that's t
14:14 piontec *it
14:14 salmon_ mattymo: why do you want store password on nodes?
14:14 mattymo it's how we deploy currently
14:14 piontec .... (a few hours wasted just beacuse of browser's cache... perfect)
14:15 vkramskikh piontec: thanks for your efforts, finally we discovered the culprit :)
14:15 vkramskikh that's actually the issue, we will try to fix it by 5.1
14:16 salmon_ mattymo: but we are talking about password to fuel api
14:16 piontec vkramskikh: thanks for help. when do you plan to release 5.1?
14:16 mattymo I know
14:16 piontec and i think it's worth putting somewhere in docs; maybe it wil svae some time for other people
14:17 mattymo I think you'll eventually get to this point where we can discuss again
14:17 salmon_ mattymo: ?
14:17 mattymo I think you need it in astute.yaml on master node docker containers (not deployed nodes)
14:18 vkramskikh piontec: the target date is july 31
14:19 salmon_ mattymo: ah ok, sorry
14:20 salmon_ mattymo: yes, on master node we need to store keystone service token.
14:20 mattymo that yaml defines ip of master node, which interface we're using, dns info, and dhcp range
14:21 salmon_ but this token is unchangeable
14:24 mattymo that's okay
14:30 EeeBotu New bug #1324954 in fuel: "Ceph volume cannot be attached/created in case of different cinder rbd_user and pool name" [Undecided, New] https://bugs.launchpad.net/bugs/1324954
14:30 EeeBotu New bug #1324951 in fuel: "Single network failure will break the cluster" [Undecided, New] https://bugs.launchpad.net/bugs/1324951
14:40 e0ne joined #fuel-dev
14:45 teran joined #fuel-dev
14:48 e0ne joined #fuel-dev
15:08 angdraug joined #fuel-dev
15:24 e0ne_ joined #fuel-dev
15:52 EeeBotu New bug #1324976 in fuel: "[System Tests] Cherry pick fixes for system tests 4.1.1" [High, Confirmed] https://bugs.launchpad.net/bugs/1324976
15:53 asledzinskiy left #fuel-dev
16:02 IvanBerezovskiy left #fuel-dev
16:03 Tatyanka_Leontov left #fuel-dev
16:19 r0mikiam_ joined #fuel-dev
16:21 rmoe joined #fuel-dev
16:24 r0mikiam_ joined #fuel-dev
16:51 teran joined #fuel-dev
16:53 EeeBotu New bug #1325012 in fuel: "UI is cached between Fuel versions" [High, Confirmed] https://bugs.launchpad.net/bugs/1325012
16:58 MiroslavAnashkin mihgen: Mike, what are the latest rumors about HP P400 controllers we wanted to order in US on eBay?
17:07 mihgen MiroslavAnashkin: yes, I want to make it happen
17:08 mihgen if you help me and provide eBay links, I'll be more than happy and will do it sooner :)
17:08 mihgen christop1eraedo: hi Chris
17:08 mihgen did you create a bug about ISO build script by any chance.. ?
17:24 IlyaE joined #fuel-dev
17:25 xarses joined #fuel-dev
17:39 e0ne joined #fuel-dev
17:43 e0ne_ joined #fuel-dev
18:06 Arminder joined #fuel-dev
18:09 IlyaE joined #fuel-dev
18:10 christop1eraedo mihgen: yes
18:11 christop1eraedo Had a conversation with Roman, Mike and Andrew about improving the process for doing Fuel development
18:11 dshulyak_ joined #fuel-dev
18:11 MiroslavAnashkin vkramskikh: Should we add message to our Fuel UI first screen, pointing customer to perform Shift+Reload page in order to update browser cache?
18:34 testmaniac trying to get an environment created in fuel 5.0.... Only the controller will get through the setup process, the compute node and storage nodes (ceph) have errors starting with: (ceph-deploy --overwrite-conf config pull node-11 returned 1 instead of one of [0]) then a cascade of other puppet errors.
18:34 testmaniac what is the best way to troubleshoot this and report to you guys?
18:35 angdraug testmaniac: 9 times out of 10 failure of ceph-deploy config pull means management network between the nodes is broken
18:35 angdraug see if ssh node-11 works
18:36 angdraug also check if node-11 has /etc/ceph/ceph.conf
18:55 testmaniac angdraug: So the "Verify Network" would not catch that the management network was broken?
18:56 testmaniac i can ssh into node-11 fine, and there is a /etc/ceph/ceph.conf
18:58 e0ne joined #fuel-dev
18:59 angdraug what happens if you run "ceph-deploy --overwrite-conf config pull node-11" manually?
18:59 angdraug anything interesting in /root/ceph.log?
19:02 testmaniac 2014-05-30 17:48:30,896 [node-11][INFO  ] write cluster configuration to /etc/ceph/{cluster}.conf 2014-05-30 17:48:30,938 [node-11][ERROR ] Traceback (most recent call last): 2014-05-30 17:48:30,941 [node-11][ERROR ]   File "/usr/lib/python2.6/site-packages/ceph_deploy/util/decorators.py", line 10, in inner 2014-05-30 17:48:30,943 [node-11][ERROR ]     def inner(*args, **kwargs): 2014-05-30 17:48:30,945 [node-11][ERROR ]   File
19:03 testmaniac running manually seems ok
19:03 angdraug hm, that's not very telling :(
19:03 angdraug if running manually is ok, this may be a timing issue
19:03 angdraug config pull gets triggered too early
19:06 e0ne joined #fuel-dev
19:06 testmaniac is there a workaround?
19:09 testmaniac when running /usr/bin/ceph-deploy --overwrite-conf config pull node-11 from node 12, i get an error
19:09 testmaniac failed to fetch
19:10 testmaniac maybe that does point to networking... it is all VLAN, setup the same way it was when using fuel 4.1 last week
19:10 angdraug wait, I'm confused
19:11 angdraug are you still getting an error when running from node-12, or was that only during deployment?
19:12 testmaniac well, i tested the ceph-deploy --overwrite-conf from node11 first... it was pulling it from itself
19:12 testmaniac so it worked
19:13 testmaniac then i sshed into node12
19:13 testmaniac and tried the ceph-deploy command again... this time i got an error
19:14 testmaniac let me check that the vlans are numbered correctly
19:14 testmaniac just in case
19:17 testmaniac doing an env-reset to try again, re-numbered some vlans to match identically the fuel 4.1 setup i had before
19:38 EeeBotu New bug #1325071 in fuel: "global name 'out_path' is not defined" [Undecided, New] https://bugs.launchpad.net/bugs/1325071
19:38 EeeBotu New bug #1325068 in fuel: "Master node installation should not delete existing partitions on target disk silently" [Undecided, New] https://bugs.launchpad.net/bugs/1325068
19:44 IlyaE joined #fuel-dev
20:18 casanch1 joined #fuel-dev
20:27 e0ne joined #fuel-dev
20:52 e0ne joined #fuel-dev
21:32 vkramskikh MiroslavAnashkin: we shouldn't, we should fix it properly. I've filed a bug for 5.1: https://bugs.launchpad.net/fuel/+bug/1325012
21:36 IlyaE joined #fuel-dev
22:23 xarses joined #fuel-dev
22:50 teran joined #fuel-dev
23:01 christop1eraedo try to create a clean fuel dev environment, seems like npm doesn't get installed before it's needed.  (step 3.4 http://docs.mirantis.com/fuel-dev/develop/env.html)
23:01 christop1eraedo is it safe to "apt-get install npm" before "npm install -g grunt-cli"
23:03 christop1eraedo Also, I'm attempting to skip directly to step 3.4, which should work (and is implied by step 3.4.3), so I have not run through nailgun development steps at http://docs.mirantis.com/fuel-dev/develop/nailgun/development/env.html
23:18 christop1eraedo cd ../fuel-main
23:35 salmon_ joined #fuel-dev

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