14:00:25 <saggi> #startmeeting smaug
14:00:26 <openstack> Meeting started Tue Apr 19 14:00:25 2016 UTC and is due to finish in 60 minutes.  The chair is saggi. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:00:28 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:00:30 <openstack> The meeting name has been set to 'smaug'
14:00:34 <saggi> Hi everyone
14:00:37 <xiangxinyong456> hi
14:00:40 <yinweimac> hi
14:00:43 <Guest45645> hi
14:00:48 <yuval> hey
14:01:08 <chenying_> hi
14:01:15 <zhonghua-lee> hi
14:01:19 <saggi> gampel is away on holiday. He's collection energy for the summit
14:01:30 <saggi> *collecting
14:01:36 <yinweimac> cool
14:01:55 <xiangxinyong456> :)
14:02:13 <Guest45645> how much has he collected for now?:)
14:02:13 <saggi> Is everyone here?
14:03:06 <zengchen> yes
14:03:09 <saggi> good
14:03:23 <saggi> #topic OpenStack Summit
14:03:29 <saggi> Summit is next week
14:04:01 <saggi> I'm working on a presentation for the vbrownbag. We get 15 minutes.
14:04:16 <saggi> I'll send everyone a copy of it later today so you can comment.
14:04:25 <xiangxinyong456> good
14:04:34 <zengchen> have a good trip! great!
14:04:38 <yinweimac> sure
14:04:43 <chenying_> Ok
14:04:44 <saggi> It's supposed to just introduce Smaug and point out our main advantages
14:05:01 <saggi> Mainly our pluggable design. And the whole project attitude.
14:05:08 <saggi> These are the things that set us apart
14:05:34 <xiangxinyong456> saggi: eran told us you need a video
14:05:44 <saggi> We would also like to have some sort of a demo video. xiangxinyong456, did gampel speak with you about it?
14:05:47 <yinweimac> yes, I used to be asked what's the difference with project freeze
14:06:06 <xiangxinyong456> yeah
14:06:30 <saggi> yinweimac: I'm also trying to get a hold of the freezer guys to see how we can work together instead of compete.
14:06:38 <saggi> xiangxinyong456: How is that going?
14:06:52 <xiangxinyong456> firstly i will send the video to mailinglist
14:07:44 <xiangxinyong456> i plan to record the video tomorrow
14:08:10 <xiangxinyong456> but i have no idea about the operation logs
14:08:49 <saggi> xiangxinyong456: What do you mean?
14:09:48 <xiangxinyong456> because the operation logs have something undecided
14:10:23 <saggi> xiangxinyong456: Well talk about it offline. In either case we might just skip showing it in the video. It's just a short demonstration anyway.
14:10:45 <xiangxinyong456> ok.
14:10:57 <chenying_> Now the operation log do not cover the scene portecting immediately without scheduler.
14:10:58 <zhonghua-lee> is there another topic about that?
14:11:22 <saggi> zhonghua-lee: About what?
14:11:33 <xiangxinyong456> operation logs
14:11:46 <xiangxinyong456> i add it into wiki
14:11:56 <zhonghua-lee> saggi: operation logs
14:12:13 <saggi> no but I'll add one
14:12:14 <saggi> :)
14:12:32 <yuval> chenying_: and do you think they should?
14:12:32 <saggi> Any other comments about summit?
14:13:09 <xiangxinyong> saggi, i will send a video to you tommorrow night
14:13:24 <saggi> xiangxinyong, great!
14:13:31 <xiangxinyong> by email?
14:13:39 <saggi> yes
14:13:43 <saggi> if it fits
14:13:47 <xiangxinyong> ok
14:14:06 <xiangxinyong> or i can upload it to a website
14:14:29 <zhonghua-lee> I wish I could attend the summit.:(
14:14:32 <chenying_> yuval: IMO, we should see the progress and status of all operation type from the operation log API .
14:14:52 <saggi> xiangxinyong, I wouldn't like WIP versions of the video public. So if you upload it anywhere make it private.
14:14:59 <xiangxinyong> chenying_: i agree with you.
14:15:06 <saggi> #topic operation logs
14:15:06 <yuval> chenying_: I totally agree. Problem is, there might not be an operation engine running, only a protection service
14:15:15 <xiangxinyong> saggi: ok
14:15:41 <chenying_> checkpoint
14:16:08 <saggi> The problem with having operation logs without an operation is that the information will have to go in the bank if it's checkpoint related.
14:16:11 <chenying_> all protect action will call create_checkpoint
14:16:23 <chenying_> create_checkpoint rest API
14:16:50 <saggi> This is problematic as it breaks the CRUD nature of REST
14:17:08 <saggi> As it is now, information about checkpoint progress is inside the checkpoint.
14:17:20 <saggi> We can add more information to the checkpoint, like logs.
14:17:31 <saggi> But this will have different endpoints.
14:18:02 <chenying_> so we can get the progress and status of all operation from  checkpoint  information?
14:18:26 <saggi> The point is we can't mix those up.
14:18:42 <saggi> We want to support the user case of a user never scheduling operations
14:18:56 <saggi> having it's own management logic somewhere and only using the simple operations.
14:19:04 <xiangxinyong> saggi: but we can not distinguish the direct protect and schedule protect from the checkpoints
14:19:10 <saggi> This means that the information needs to be part of the entity being created.
14:19:35 <saggi> It Shouldn't matter how the checkpoint was created for the restore side
14:20:27 <saggi> The main issue is whether having logs for the checkpoint is really necessary.
14:20:35 <saggi> We currently provide information about current progress.
14:20:40 <xiangxinyong> saggi: we should see the checkpoint info from the checkpoint page
14:20:45 <saggi> So there are no logs.
14:21:13 <saggi> Meaning there is no infiltration about *when* things changed. Just what the status is now.
14:21:30 <saggi> I want to know if knowing when things changed really makes sense in the context of a checkpoint.
14:21:47 <saggi> How does cinder\nova do per instance logs
14:21:50 <saggi> do they even?
14:22:21 <xiangxinyong> sagg: do you mean we only need the restore to show in the operation logs page?
14:22:30 <chenying_> From xinyong's opinion, He want a or several API that can show the progress and status of a operation(protect delete restore) in smaug ui.
14:22:51 <saggi> We have progress status from the checkpoint already
14:23:19 <chenying_> How does cinder\nova do per instance logs   They will not got to a new page, just show the status of the resoure like volume.
14:23:26 <saggi> When you ask for information about a checkpoint you should get progress information (if it makes sense).
14:23:32 <chenying_> resource
14:24:24 <xiangxinyong> how could we show the delete workflow when we delete a checkpoint?
14:24:59 <saggi> What do you want to show?
14:25:02 <yinweimac> we just show the checkpoint is deleting
14:26:07 <xiangxinyong> so we could change the name "operation logs" to "Restores"?
14:26:09 <xiangxinyong> ?
14:26:10 <xiangxinyong> ?
14:26:14 <chenying_> Un cinder the the progress and status of volume is recorded in the resoure(volume) db table, will not create a new api or db table to record the  data log.
14:26:22 <chenying_> In cinder
14:27:06 <saggi> Operation logs are for the operations. Operation could contain multiple atomic operations.
14:27:15 <saggi> *atomic actions
14:27:32 <saggi> So operation logs "link" to object
14:28:09 <xiangxinyong> saggi: i got some confused.
14:28:57 <saggi> I'll try and do a writeup about how operations relate to other entities and than we'll try and think about how to integrate everything together once we all understand the requirements.
14:28:58 <saggi> OK?
14:29:18 <chenying_> sure
14:29:48 <xiangxinyong> ok
14:30:01 <saggi> #topic operation workflow
14:30:15 <saggi> #topic operation workflow
14:30:30 <saggi> OK
14:30:37 <yuval> As you might have seen, I sent a mail to the openstack-dev (tagged smaug) with a suggestion for the restore/protect resource workflow
14:31:49 <yuval> It basically answer the need for protection plugins to perform work both independently of child resources and other work which depends on child resources
14:33:26 <saggi> Did everyone read it?
14:33:41 <chenying_> I have seen. I will take some time to read it.
14:33:50 <saggi> yinweimac?
14:33:55 <yinweimac> yes
14:34:08 <yinweimac> actually I have a talk with yuval already
14:34:30 <yinweimac> I agree to divide sync/paralle tasks
14:35:15 <yinweimac> but at the same time, I have a bit concern that it looks complicating to protection plugin developers
14:35:54 <saggi> Yuval and I were talking about it this morning and identified that the main issue is that TaskFlow doesn't have any support for async tasks.
14:35:55 <yinweimac> and I also gave one fs consistency case to yuval, to see how to meet this case in his design
14:36:27 <yinweimac> not actually
14:36:32 <yuval> yinweimac: what is that fs consistency case?
14:36:56 <yinweimac> server freeze io before volumes to take snapshot
14:37:48 <yuval> yinweimac: let me think about it, although that might be an issue with the current workflow patches as well, right?
14:37:57 <yinweimac> yes
14:38:14 <yinweimac> I think we may need take time to think more about the task flow
14:38:15 <yuval> ok, so let's take it offline after the meeting
14:38:43 <yinweimac> there's a trade off between the complexity and the clear semantics
14:38:44 <yuval> I suggest that everyone should read the mail and the attached pdf, so we will have more eyes and opinions
14:39:01 <saggi> #topic open issues
14:39:01 <yuval> *eyes reading it
14:39:04 <yinweimac> but I'm busy with the integration demo
14:39:11 <yinweimac> right now
14:39:53 <yinweimac> one suggestion is, how about have smaug all services integrated and make them work first?
14:40:15 <yinweimac> then we start an enhance iteration?
14:40:32 <saggi> yinweimac: enhance iteration?
14:40:47 <saggi> oh
14:40:48 <saggi> sure
14:41:00 <saggi> yinweimac: Yes, I understand that the demo is high priority
14:41:06 <yuval> Open issue: I think that we should share on some tasks site (waffle? trello?) the open tasks that people are working on. I finished working on an issue, only to find out that someone already submitted a patch for the same issue.
14:41:40 <saggi> We could use trello, dragonflow uses it and they recommend it
14:41:42 <zengchen> i have one patch need to be merged, otherwise the OperationEngine can not run. can everyone take you time to review it. it stays for a long time. thanks. https://review.openstack.org/#/c/282263
14:41:51 <yinweimac> hmm, eran used to have epad to assign tasks
14:42:11 <saggi> Yes,but we found it hard to maintain.
14:42:21 <yinweimac> report bug or bp is an explicit way
14:42:46 <chenying_> I agree with yinwei.
14:42:54 <saggi> I don't mind using bugs.
14:43:10 <saggi> But that means that everyone needs to open bugs for the tasks they are working on (if no bug exists)
14:43:13 <chenying_> Can record the task using bug or bp.
14:43:21 <saggi> And tag them as enhancement if they are not actual bugs.
14:43:43 <yinweimac> enhancement could be bp?
14:44:18 <yinweimac> btw, will the bugs on launch pad be closed automatically if the patch associated with it has been merged?
14:44:19 <saggi> bp is it's own kind of headache.
14:44:28 <saggi> yinweimac: They should
14:44:28 <yuval> yinweimac: I think so
14:44:51 <yuval> the launchpad bugs/bp system is ok, but I actually think a proper task site is much more convinient
14:45:17 <yinweimac> ok, let's have a try with trello
14:45:21 <yuval> but I don't mind either, as long as we really use it
14:45:49 <yinweimac> both will do for me
14:46:16 <saggi> Do you guys have access to trello from china?
14:46:31 <xiangxinyong> it is ok
14:46:54 <xiangxinyong> we used it in the last year
14:47:35 <saggi> OK
14:47:46 <saggi> Any other topics?
14:48:38 <yuval> Umm
14:48:50 <yuval> There are some old patches which require some reviews
14:49:13 <yuval> and some other which are on -1 validation and require work
14:49:41 <chenying_> I have one . When will we start to design and develop the backup and replication  plug-in?
14:50:15 <saggi> We are working on backup now
14:50:17 <yinweimac> backup plugin have been developed already
14:50:35 <chenying_> sorry It is snapshot
14:50:39 <saggi> replication will have to wait after we have regular backup working end-to-end
14:50:46 <saggi> snapshots as well
14:50:57 <saggi> I'd very much like to get the simple things done first
14:51:04 <chenying_> Ok
14:51:07 <saggi> #action (sagg) Prepare document describing logging issue
14:51:25 <saggi> just so I don't forget
14:51:29 <yuval> I will open the Trello board and send the link in #openstack-smaug
14:51:30 <xiangxinyong> Thanks saggi.
14:52:04 <saggi> #action (yuval) open trello board
14:52:23 <saggi> Any other action items I might have forgot?
14:53:01 <saggi> Ok, than I think we're done
14:53:01 <saggi> Thanks everybody
14:53:14 <yinweimac> thans
14:53:17 <yinweimac> thanks
14:53:17 <xiangxinyong> Thanks
14:53:19 <yinweimac> bye
14:53:19 <chenying_> thanks  all
14:53:21 <yuval> thanks, bye
14:53:23 <xiangxinyong> Bye
14:54:04 <saggi> #endmeeting