08:00:22 <ifat_afek> #startmeeting vitrage
08:00:23 <openstack> Meeting started Wed Nov 28 08:00:22 2018 UTC and is due to finish in 60 minutes.  The chair is ifat_afek. Information about MeetBot at http://wiki.debian.org/MeetBot.
08:00:24 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
08:00:26 <openstack> The meeting name has been set to 'vitrage'
08:00:30 <ifat_afek> Hi :-)
08:00:31 <eyalb> \o/
08:01:35 <e0ne> hi
08:03:07 <ifat_afek> #topic Status and updates
08:03:14 <ifat_afek> No special updates today
08:03:20 <ifat_afek> I’m working on bugs and also progressing with the Nova versioned notifications.
08:03:31 <ifat_afek> Besides that, Alon and I tried to help with e0ne’s efforts to use xstatic packages in vitrage-dashboard.
08:03:33 <idan_hefetz> Hi :)
08:03:35 <ifat_afek> That’s it for me
08:04:01 <e0ne> ifat_afek: thanks for the help. I hope to get more time for this stuff this week
08:04:13 <ifat_afek> e0ne: thanks for your help
08:04:23 <e0ne> I'm working on the tempest plugin refactoring too
08:04:27 <eyalb> ifat: any progress about that ?
08:04:38 <ifat_afek> eyalb: about the xstatic?
08:04:42 <eyalb> yes
08:04:54 <e0ne> eyalb: I've got working PoC now
08:04:59 <ifat_afek> Alon and I made it work, but there are some parts we don’t fully understand. So we asked for e0ne’s help
08:05:10 <e0ne> eyalb: but it didn't work with compressed static :(
08:05:19 <ifat_afek> I mean, Alon and I finished wht e0ne started :-)
08:05:29 <eyalb> eOne: do you know why ?
08:05:30 <e0ne> ifat_afek: :)
08:05:37 <e0ne> eyalb: not yet
08:05:59 <eyalb> eOne: whats the error ?
08:06:01 <e0ne> TBH, I don't have enough time to dig into this
08:06:07 <ifat_afek> Now compress works, but only if a flag is False, and with some import lines that we are not sure why they are needed
08:06:35 <ifat_afek> And I’m speaking on behalf of Alon, I was mainly watching what he did ;-)
08:06:38 <eyalb> if the flag is False it meanes not compress no ?
08:07:10 <ifat_afek> Yes, but what matters is the value of the flag when you run compress. Compress works anyway, but if you run it with a flag True then later on nothing works
08:07:22 <ifat_afek> If you run compress with False and then change it to True, it works well
08:07:25 <e0ne> eyalb: I don't remember which errors I faced
08:07:35 <e0ne> I can share a PoC to the openstack-dev@ if needed
08:07:52 <ifat_afek> e0ne: we didn’t face any error. But we started by testing without compress, fixed some issues, and only then tried to compress
08:08:03 <e0ne> ifat_afek: hm.. interesting. I'll taka a look o it asap
08:08:27 <ifat_afek> e0ne: when you have the time, we can wait
08:08:44 <e0ne> ifat_afek: what did you fix?
08:09:46 <ifat_afek> You removed some scripts from the index.html. In alarms index.html we added the scripts back, but with {{STATIC_URL}} prefix
08:09:54 <ifat_afek> It solved the missing dagre
08:10:12 <ifat_afek> Which was needed for the RCA view
08:10:20 <e0ne> ifat_afek: ok, thanks. I'll take a lookk on it
08:10:41 <ifat_afek> Thanks!
08:10:43 <e0ne> ifat_afek: I want to finish  this task sooner to start a new one
08:11:19 <ifat_afek> Sooner is always good :-) and thanks again for your help with that, because we don’t have enough knowledge in this area
08:11:45 <ifat_afek> BTW we sent you an email with the exact changes we’ve made
08:12:29 <e0ne> I was o PTO yesterday, didn't read the mail yet
08:13:02 <ifat_afek> OK
08:13:40 <idan_hefetz> Currently trying to improve the run time and memory usage of our APIs.
08:13:45 <idan_hefetz> Mainly topology which is the most wasteful but also resource list.
08:13:55 <idan_hefetz> Seems there is at least one unneeded graph.copy() that can be removed while changing the algorithm a bit.
08:14:04 <idan_hefetz> Also, eyalb suggested to compress the data sent via RPC between the api handler and the api.
08:14:21 <idan_hefetz> Fixing these two issues and testing on a 100,000 vertices graph, shows runtime is reduced to about a half.
08:14:29 <ifat_afek> Wow
08:14:59 <idan_hefetz> Hope to commit these changes soon
08:15:00 <eyalb> we tryed using gzip for the http transport but there wasn't much improvment
08:15:33 <eyalb> although we saw that the http content was reduced
08:15:38 <ifat_afek> So most of the time is spent on the sub-graph queries?
08:15:48 <eyalb> i guess that it is good only for gigs of info
08:15:50 <ifat_afek> I mean graph queries
08:15:51 <idan_hefetz> mabey if the network is slow it will be more noticable
08:16:30 <eyalb> i saw that glance added this option for there images but it is not used by default
08:16:54 <idan_hefetz> ifat_afek: no actually most of the time is spent copying the graph to provide a result that is different from the original graph
08:17:10 <idan_hefetz> and also spent outpuing it to a json
08:17:29 <ifat_afek> Ok, thanks
08:17:46 <idan_hefetz> eyalb: this change can be added to the configuration, so we can document it
08:18:32 <eyalb> we can add it is just a config
08:18:59 <e0ne> idan_hefetz, eyalb: it would be great to see some profiling results
08:20:58 <idan_hefetz> sure i'll try to create something documented, please send an example if you have something in mind
08:22:48 <e0ne> sounds good
08:23:28 <ifat_afek> Any other issue for today’s meeting?
08:24:45 <ifat_afek> See you next week!
08:24:55 <eyalb> bye
08:25:19 <ifat_afek> #endmeeting