13:01:28 <zhipeng> #startmeeting tricircle
13:01:29 <openstack> Meeting started Wed Sep  9 13:01:28 2015 UTC and is due to finish in 60 minutes.  The chair is zhipeng. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:30 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:33 <openstack> The meeting name has been set to 'tricircle'
13:01:42 <zhipeng> #topic Roll Call
13:01:50 <zhipeng> hello folks
13:01:58 <gampel> #info gampel
13:02:05 <zhiyuan_> #info zhiyuan
13:02:06 <zhipeng> #info zhipeng
13:02:14 <zhipeng> i didn't see joe here
13:02:56 <zhipeng> #topic work progress and discussion
13:02:59 <saggi> #info sggi
13:03:04 <saggi> #info saggi
13:03:22 <zhipeng> #info joehuang
13:03:26 <joehuang> hello
13:04:20 <saggi> hi
13:04:29 <joehuang> hi saggi
13:04:30 <gampel> hi
13:04:36 <joehuang> hi gampel
13:05:24 <joehuang> the topic is work progress, right
13:05:35 <zhiyuan_> yes
13:05:37 <saggi> yes
13:05:48 <gampel> zhiyuan do you want to start ?
13:06:18 <joehuang> zhiyuan first?
13:06:24 <zhiyuan_> fine
13:06:45 <zhiyuan_> I have just submitted the new patch for client wrapper.
13:07:26 <joehuang> good, I will find time to review it
13:08:26 <saggi> me too
13:08:29 <joehuang> for me. I update the rest api framework with devtstack plugin
13:08:31 <zhiyuan_> most of the comments are addressed. and I remove all the explicit declaration of list_(resource)s method
13:09:01 <joehuang> and receive some comments on it
13:09:10 <gampel> does it include API to get the urls of the bottom stack ?
13:09:32 <zhiyuan_> make them to be registered automatically by Client class, so we don't need to change Client when  we add new type of resources
13:09:33 <joehuang> not yet, it's framework to integrate keystone
13:09:52 <zhiyuan_> yes, I add a new get_endpoint method
13:11:33 <joehuang> I'll have a two week business trip to meet with customer, may find no time to update the patch of the rest framework, including API these two weeks
13:11:47 <joehuang> but I'll try to review patches you submitted
13:12:24 <gampel> Ok zhiyuan saggi are we dependent for the next task on Joe patch ?
13:12:34 <zhiyuan_> any new requirement for DAL so I can continue to work?
13:12:47 <saggi> I would like to get the WSGI up and running but I'm not blocked on it.
13:13:03 <gampel> yes saggi compiled a list of DAL tasks
13:13:05 <joehuang> I suggest you first insert some record into db tables if you need it
13:13:16 <zhiyuan_> I can update the patch of client and start a new patch for new requirement at the same time
13:13:25 <gampel> saggi maybe you could update us on the status first
13:13:30 <saggi> Sure
13:14:10 <saggi> I sent an update to the nova skeleton. This ties everything in the adapter side. It also fixes some bugs.
13:14:27 <saggi> I also sent a patch that breaks all other patches
13:14:44 <saggi> It moves all the adapter code to it's own service and leaves the cascade serivce blank
13:15:06 <saggi> It's just so that people can base their work on it.
13:15:26 <saggi> And because it's very hard to review patches that contain file moves *and* modifications.
13:15:36 <saggi> So it only contains moves and minimal modifications.
13:16:16 <saggi> I'm now working on getting versioned objects so that we can communicate across the message queue and DB
13:17:09 <saggi> it's shou;dn
13:17:30 <saggi> it shouldn't effect DAL or WSGI work
13:17:36 <saggi> since they use different files
13:18:01 <joehuang2> my connection lost for a while
13:18:25 <saggi> But I would like to push the changes I have up so that we have a common master to work from
13:18:28 <saggi> So please review
13:18:31 <joehuang2> would the google doc be updated these days to reflect some new idea
13:18:40 <saggi> Yes, gampel is working on it
13:18:46 <gampel> I am working on the design document with all the development and scalability feedback i think it will be ready By monday morning
13:19:04 <joehuang2> ok. I will also review the code ASAP
13:19:31 <gampel> lets all try to review  zhiyuan and saggi patch tomorrow
13:19:38 <gampel> so we could merge it all in
13:19:38 <joehuang2> yes
13:19:45 <joehuang2> agree
13:19:58 <saggi> In case you wonder what supposed to work in my nova patches.
13:20:15 <saggi> You should be able to create a VM and it will finish scheduling and appear as active
13:20:20 <saggi> but fail due to the port not being up
13:20:29 <saggi> since the network bits are not wrapped yet
13:21:08 <joehuang2> got it
13:21:28 <zhiyuan_> my connection also lost for a while
13:21:40 <gampel> Saggi can you please share your ideas for the next DAL tasks
13:21:51 <gampel> and overall tasks
13:22:18 <saggi> The thing I really wanted was WSGI <-> DAL tied down
13:22:28 <saggi> but that will have to wait.
13:22:56 <saggi> Another things that needs to be done is syncing the AZ between tricircle and nova
13:23:09 <saggi> currently I manually add the fake compute nodes to different AZs
13:23:46 <saggi> I don't really know how to block the user from removing a compute node from an AZ
13:24:12 <saggi> but we need to make sure that the AZ we define in the tricircle DB is the same that is defined in nova
13:24:16 <joehuang2> you can add some fix mapping first to adding code for other parts and test it works or not
13:24:18 <saggi> and that it's always defined there
13:25:21 <saggi> So in general, the task is: Make sure that the compute node representing the site is in an AZ we define and nova knows about that mapping.
13:26:22 <joehuang2> when you create a fake node, you can add it to the az you want in the nova db in the top layer
13:26:37 <gampel> if i understand correctly   final flow should  be add-site --> WSGI --> DAL --> NOVA
13:26:47 <saggi> YES
13:26:59 <saggi> WSGI --> Cascade Service --> DAL --> NOVA
13:27:25 <joehuang2> but api can't see the fake node, it's dynamicly created
13:27:51 <saggi> Once you create the fake node in the Adapter it registers with the DB so top knows about it
13:27:58 <saggi> The flow would be:
13:29:10 <zhiyuan_> so DAL needs to register new site to the top NOVA, as compute node?
13:29:25 <gampel> yes i think so
13:29:32 <gampel> as a AZ
13:29:51 <saggi> User -create_site()-> WSGI -create_site()-> Cascade Service -create_site()-> DB
13:29:51 <saggi> -create_fake_node-> Adapter
13:29:51 <saggi> -add_to_az()-> Dal --> Nova
13:30:11 <saggi> Trying to do a sequence diagram in ASCII is hard :)
13:30:23 <joehuang2> one moment
13:30:41 <gampel> :)
13:31:19 <zhipeng> new occupation as ASCII artist :P
13:31:35 <joehuang2> if no compute node, can the AZ record be created directly in Nova DB
13:31:55 <saggi> We don't want to access the Nova DB directly
13:32:09 <gampel> it will be better to do it via the API
13:32:21 <saggi> Having the function that bind the AZ to a Compute host would suffice for now. Since I need to move some code around have the sites defined by the DB and no an in memory list
13:32:47 <joehuang2> it seems that the AZ is prensented only when the compute node was added into host aggregate, in current nova implementation
13:32:55 <saggi> yes
13:33:00 <saggi> it's just a string attribute
13:33:05 <joehuang2> create_site()-> WSGI -create_site()-> Cascade Service -create_site()-> DB, there is no compute node in the AZ yet. I am affraid can we insert a new AZ into NOVA
13:33:14 <saggi> yes
13:33:22 <saggi> the second phase is creating a node in the adapter
13:33:28 <saggi> that will register with nova
13:33:38 <saggi> This is what my current patches implement
13:34:15 <saggi> Than after it registered with nova add it to a Host Aggregate with the proper AZ attribute
13:34:37 <joehuang2> create_site()-> WSGI -create_site()-> Cascade Service -create_site()-> DB->create fake node->create host aggreagte in nova
13:35:00 <saggi> That is the order
13:36:09 <joehuang2> :)
13:37:06 <joehuang2> I think work progress and status is clear, may we end the meeting earlier
13:37:20 <zhiyuan_> sorry I lost the connection again and I change the client. Do I miss something?
13:37:53 <zhiyuan_> I only see the order sent by joehuang
13:37:58 <gampel> welcome back we think that the next DAL task is handle site registration
13:38:03 <zhiyuan_> this "create_site()-> WSGI -create_site()-> Cascade Service -create_site()-"
13:38:18 <saggi> http://www.plantuml.com/plantuml/img/TP3DIaD138NtUOfmztq15zAV2wKBNRGKmTtfJ9mEtdkdJd8hlhqP196gsvplV0bPgg7QF0v4bcqGxXv45KcK5toGjrcc8Dndf8gn8qAmKlceKkbKvYcYVl3JORiZQZDUtFFskECx3bLWmffDg4qyw3RGW2YkhTSSfBl18Wmdg6zRqs_M1Vtwfl4djOguc9_oZRxWLJYKyJB_L-anEBwNbBKrcvlVivrvzKmjS-InNE7CvajvBcgCbAea_61fALFqNtu2
13:38:23 <joehuang2> create_site()-> WSGI -create_site()-> Cascade Service -create_site()-> DB->create fake node->create host aggreagte in nova , That is the order
13:39:04 <zhiyuan_> ok, i see the figure
13:39:12 <joehuang2> good
13:39:13 <saggi> Fixed small error
13:39:13 <saggi> http://www.plantuml.com/plantuml/img/TP2zQWD138HxFOMutY-G8lZd8AH8YXXZI3VUbTTBxcwTbSx1Rs-T0o4uSQcPRqP2SpLK6_kEoB9rmisBg2A9yepNyZdA48JRZ5JHDqG8LYflL2hjoZX4r3Dlrq_FH9F6iyVhn0ySgi24DPlGLNAhMK43edXnFUKWpGqNOTX1VT-I_h9Uq2vl4lzIYuYZ-J4_w1uVmg7qn_5Uf4Nd-9kah5DowkQlR0TUlDFaEV9QJd1czLrvA6gCbAea_69fBaFqRru0
13:39:54 <saggi> Speed UML drawing :)
13:41:26 <saggi> I might also add a hook in the scheduler since it get's notification when a host aggregate changes to tell the Cascade Service to make sure everything is correct
13:43:14 <saggi> That way an admin can't screw up the mapping
13:43:35 <joehuang2> only admin can do that
13:44:58 <gampel> zhiyuan we can discuss it some more tomorrow in the tricircle IRC after you look at it a bit
13:45:01 <saggi> But an admin that doesn't know that s\he should change the AZ name in the tricircle API and not the Nova API might cause the system to misbehave
13:45:03 <saggi> better be save
13:45:37 <zhiyuan_> @gampel ok
13:48:45 <gampel> I think that we are all done thank you lets continue talking  this week in the tricircle IRC
13:49:14 <saggi> Just to make it clear
13:49:26 <saggi> Start with the dal -> nova part
13:49:37 <saggi> since I am making a lot of changes all around the cascade service
13:49:55 <saggi> don't want to have issues merging.
13:52:29 <zhipeng> so let's conclude today's meeting
13:52:36 <zhipeng> if there is no other buisness
13:52:55 <gampel> i don't think so  good night
13:53:13 <zhiyuan_> thanks and bye~~
13:53:18 <joehuang2> bye
13:53:26 <zhipeng> #endmeeting