17:34:54 <nati_ueno> #startmeeting quantum-tempest
17:34:55 <openstack> Meeting started Thu Sep 20 17:34:54 2012 UTC.  The chair is nati_ueno. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:34:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
17:34:57 <nati_ueno> is this works?
17:34:58 <openstack> The meeting name has been set to 'quantum_tempest'
17:35:13 <nati_ueno> oh its works! This is useful
17:35:36 <nati_ueno> #info Current discussion summary http://etherpad.openstack.org/quantum-tempest
17:36:24 <nati_ueno> Ok I wanna talk about testing tool until waiting zyluo's update
17:36:33 <nati_ueno> #topic testing tool
17:36:53 <nati_ueno> mnewby: I wanna hear your idea about quantum-debug command
17:40:44 <zyluo> what does that command do?
17:41:17 <zyluo> and where is it going to be located? Tempest or Quantum
17:41:18 <nati_ueno> It create port for test.
17:41:19 <nati_ueno> https://github.com/openstack/quantum/tree/master/quantum/debug
17:41:36 <zyluo> ic i'll take a look
17:41:39 <nati_ueno> We can test connectibility without VM
17:41:48 <zyluo> oh cool
17:42:28 <nati_ueno> It uses namespace function, so each testing environments are isolated
17:42:36 <nati_ueno> zyluo: Thanks
17:44:52 <nati_ueno> hmm it looks mnewby isn't here.
17:44:59 <mnewby> sorry, i'm on a call...
17:45:08 <nati_ueno> mnewby: Oh I got it :)
17:45:59 <zyluo> Then we can also extend this to test isolation, right?
17:46:21 <mnewby> nati_euno: ok, done
17:46:26 <nati_ueno> May I ask meaning of test isolation?
17:46:45 <nati_ueno> mnewby: k
17:46:51 <mnewby> can you please ask your question of me again?
17:47:15 <nati_ueno> OK In previous meeting, you said something is needed to quantum-debug command
17:47:20 <nati_ueno> So I wanna here your idea
17:47:29 <nati_ueno> to make the command more useful for testing
17:47:49 <mnewby> On the subject of namespace isolation, the current debug agent is not useful because a) it requires that tempest be running on the same box as the target and b) it requires sudo privileges to do so.
17:48:15 <mnewby> The debug agent needs to be able to run remotely if it is to be useful.
17:48:22 <nati_ueno> For a) quantum-debug command can be run remotely
17:48:26 <mnewby> Likely provided as a privileged quantum extension.
17:48:40 <nati_ueno> I agree for b)
17:48:42 <mnewby> nati_ueno: it can be called via an api?  I wasn't sure of that.
17:49:14 <nati_ueno> mnewby: Ah you mean rpc. If so it not now. But we can use ssh for testing
17:50:14 <mnewby> nati_ueno: again, that isn't really acceptable for tempest testing.
17:50:21 <mnewby> nati_ueno: ssh'ing into a vm, yes.
17:50:31 <mnewby> ssh'ing into a host, not so much.
17:50:45 <nati_ueno> mnewby: I agree it is current tempest scope
17:51:16 <nati_ueno> We should not use host ssh for blackbox testing
17:51:29 <nati_ueno> However we can have whitebox testing which has ssh for host
17:52:20 <nati_ueno> Rohit is also working stackmonkey testing tool which kills openstack or mysql processes during test
17:52:30 <nati_ueno> So IMO we can extend scope of tempest
17:52:43 <nati_ueno> We can put it on different directories
17:53:21 <nati_ueno> I agree for rpc is more useful and cool. But I feel implementing rpc is too much for testing..
17:54:09 <zyluo> makes sense to me
17:55:56 <nati_ueno> zyluo: Thanks
17:56:01 <nati_ueno> mnewby: How about you?
17:56:31 <mnewby> I have my doubts
17:56:42 <mnewby> I think we should be talking to Jay and the other tempest folks about this.
17:57:04 <mnewby> Having an unlimited scope for a project has the potential to make things very complicated indeed.
17:57:07 <mnewby> Plus, huge issue.
17:57:09 <mnewby> Huge
17:57:18 <mnewby> Metadata service doesn't work with namespaces.
17:57:20 <mnewby> (yet)
17:57:33 <mnewby> So compute is basically non-functional with namespaces enabled.
17:57:46 <mnewby> Can't do all kinds of configuration, including setting password and ssh keys.
17:57:57 <mnewby> Have you seen the bug.
17:58:09 <mnewby> ?
17:58:20 <nati_ueno> Not yet. But I agree for you.
17:58:34 <nati_ueno> May be it is too fast to test namespace which looks clearly broken
17:59:07 <nati_ueno> So I agree basic testing without namespace is more high priority now
17:59:08 <mnewby> Is there a need for the debug agent if namespaces aren't working?
17:59:18 <nati_ueno> mnewby: no need
17:59:40 <mnewby> So, moot then.
17:59:47 <mnewby> good/bad
18:00:34 <mnewby> on the upside, ssh'ing should work properly then.
18:00:43 <mnewby> without the debug agent.
18:01:32 <nati_ueno> It depends on topology
18:01:32 <mnewby> launch vm -> configure floating ip -> ssh
18:01:36 <mnewby> true
18:01:45 <nati_ueno> Ah I agree for floating ip
18:02:05 <mnewby> we'll need to ensure that devstack can be configured with l3 and an external network
18:02:17 <nati_ueno> yes
18:03:23 <mnewby> i'm wondering if we need to allow quantum to be queried as to what capabilities it is configured with...
18:03:29 <mnewby> I was asking dan that in the monday meeting.
18:03:31 <mnewby> like an extension
18:04:33 <nati_ueno> Ah that's needed
18:05:36 <nati_ueno> IMO, namespace support with nova is needed. I wanna test network connection without floating ip also in future.
18:05:48 <mnewby> namespace support with nova is needed, of course.
18:05:55 <mnewby> won't happen for folsom
18:06:02 <nati_ueno> I agree
18:06:16 <mnewby> did you see what i wrote in the recent review?
18:06:30 <nati_ueno> Which review?
18:06:53 <mnewby> The fix for non-namespaced metadata server support: https://review.openstack.org/#/c/13218/
18:07:11 <mnewby> The relevant comment:  Supporting overlapping network spaces should be considered out-of-scope for this change. The metadata service identifies which instance to return data for by the ip address of the caller. Identifying what network the instance is calling from, in a secure manner, will likely require making a proxy for the metadata service available on each network. Such a proxy would be able to securely determine the network id and pass it on to
18:07:12 <mnewby> metadata service. This would require not only creating and configuring the proxy, but also adding (presuming it doesn't already exist) an authenticated request path to the metadata service.
18:07:20 <nati_ueno> I'll take a look now
18:07:50 <mnewby> Definitely doable to support namespaces with nova…  Looks to be a lot of work, though.
18:08:13 <nati_ueno> hmm situation looks bad..
18:08:23 <nati_ueno> Let's fix it on G
18:08:53 <nati_ueno> And let's focus tempest testing without namespace
18:08:57 <nati_ueno> for folsom
18:09:14 <nati_ueno> OK I could understand mnewby's point
18:09:31 <mnewby> ok, cool
18:09:39 <zykes-> Will Metadata service not be supported on namespaced networks ?
18:09:44 <nati_ueno> Let's move zyluo's one
18:09:51 <mnewby> zykes-: Correct
18:09:59 <zykes-> eh, ooook
18:10:01 <nati_ueno> aha sorry typo
18:10:11 <zykes-> doesn't that break like lots of features ?
18:10:28 <mnewby> zykes-: Like what?
18:10:35 <nati_ueno> #topic zyluo's WIP
18:10:35 <zykes-> mnewby: cloud-init ?
18:10:53 <zykes-> isn't it the metadata service in Nova you mean ?
18:11:14 <mnewby> zykes-:  Yes. The funny thing is that the metadata service was completely broken until i added support this week.  And nobody had noticed.
18:11:19 <mnewby> #fail
18:11:38 <zykes-> so it won't be supported at all in Folsom ?
18:11:46 <zykes-> even with or without namespacing ?
18:11:50 <mnewby> It will be supported
18:11:56 <mnewby> The changes went into quantum and nova yesterday
18:12:01 <mnewby> and will be part of folsom
18:12:08 <mnewby> Non-namespaced only, though.
18:12:10 <nati_ueno> mnewby: +1
18:12:25 <zykes-> mnewby: that kinda sucks ;p
18:12:48 <nati_ueno> OK let's talk about zyluo's update
18:12:53 <nati_ueno> Ahhh he is left
18:13:00 <nati_ueno> hmm
18:13:13 <nati_ueno> OK mnewby do you have other topics?
18:13:22 <mnewby> nati_ueno: I'm afraid I have nothing to report.
18:13:37 <mnewby> I've been busy getting a working configuration (and finding bugs in the process).
18:13:41 <nati_ueno> mnewby: I got it! I have no progress also.
18:13:50 <nati_ueno> mnewby: Thanks!
18:13:52 <nati_ueno> Ok
18:13:57 <nati_ueno> #endmeeting quantum_tempest