Friday, 2016-10-14

*** openstack has joined #openstack-shade10:13
*** openstackstatus has joined #openstack-shade10:15
mordredGheRivero: that's a good question. we currently should be assuming that all ipv6 networks are external (hadn't found anyone who is ipv6 aware doing internal-only ipv6)12:12
mordredShrews: ok. I figured out the issue with the nodepool test12:27
mordredand it's an actual issue - but one I think we can ignore for a few patches12:27
mordredthe long and short of it is that nodepool uses TaskManager that runs with a single queue of operations - meaning only one task can run at a time12:28
mordredwith the patches to start moving over to rest - the explicit Task objects get queued, and then the rest task that runs inside of them also gets queued12:29
mordredbut the inner task can't run because the outer task is the active task in the queue12:29
mordredSO - nodepool isn't going to work until all of the tasks nodepool does are converted to pure-rests12:30
mordredthat is a good color on which calls to conver tnext though12:30
mordredso what I'm going to do, instead of organizing by client library (do all the tasks for a client library so we cna remove it from requirements)12:31
mordredI'll get the nodepool tasks done first - there arne't that many of them12:31
mordredthen we should be in a fairly safe place - I'm pretty sure that nobody in the world other than nodepool is actually passing in such a specific TaskManager :)12:32
*** openstackstatus has quit IRC12:43
*** openstack has joined #openstack-shade12:43
*** openstackstatus has joined #openstack-shade12:45
Shrewsmordred: then we should probably hold off on approving the stack until your theory is confirmed, yeah?13:20
mordredShrews: yes.13:23
mordredShrews: otherwise we'll get caught not being able to make emergency releases for nodepool if necessary13:24
mordredGheRivero: nice patch - one thing in it13:24
GheRiveromordred: still working on it. I just wanted it to go thorigh the gate to have a first pass13:25
*** abregman has joined #openstack-shade13:28
Shrewsmordred: https://review.openstack.org/#/c/386586/1/shade/openstackcloud.py   0 vs. 0.1 ... don't think that will make much difference. i use 0.1 that in np-builder and the cpu goes mega-hot13:29
Shrewsso i'm not convinced that is an improvement13:30
Shrewsbut i also don't think it hurts, fwiw13:32
mordredGheRivero: cool. thanks! it's looking good13:36
mordredShrews: harrumph. I wonder if there is a decent point which doesn't burn CPU but also isn't highly end-user noticable13:37
Shrewsmordred: probably the _right_ way to correct that is go back to the drawing board and consider other mechanisms (signal timers, thread conditionals, whateva). but who has time for that13:43
Shrewsmordred: also, should we consider creating a v2 branch in shade for your rest changes?13:44
Shrewsi sort of feel like that might be a good idea, but i've also only had 1 cup of coffee13:44
mordredShrews: maybe so, yeah13:47
mordredShrews: and I think the idea of releasing that as v2 - even if we still don't break backwards compat - is a good way to signal that there is a rather large change to the backend internals that one might want to test for a second13:48
mordredShrews: btw - I accelerated working on this because our centos friends reported that shade was a hardship on them because it requires all the client libs - at potentially different versions than openstack itself needs13:50
*** abregman has quit IRC13:50
*** abregman has joined #openstack-shade13:50
Shrewsyeah. everyone i've seen using centos+shade is pretty unhappy14:01
Shrewsmordred: and your 0.1 change has raised some interesting errors for some reason14:01
Shrewsfascinating14:01
*** abregman has quit IRC14:46
*** abregman has joined #openstack-shade14:53
mordredShrews: neat!16:31
*** kamsz has quit IRC16:55
*** kamsz has joined #openstack-shade16:58
mordredShrews: ok - so the failure makes some sense looking at the test ... the get after the metadata set doesn't have any concept that it could be getting old data. I'm going to remove that patch - I think the test shows us that the knock-on effects are too complex to deal with in this context17:01
mordredShrews: that said - I do think it's a worthwhile endeavor at some point to go through the entire api and make sure it behaves as expected when that wait time is not zero17:02
mordredbut not today or tomorrow17:02
*** jlk is now known as omgwtfjlk17:06
*** abregman is now known as abregman|afk17:12
*** olaph is now known as beefboymatface18:30
*** jlk has joined #openstack-shade18:48
*** jlk has quit IRC18:48
*** jlk has joined #openstack-shade18:48
*** omgwtfjlk has quit IRC18:51
*** yolanda has quit IRC19:16
*** yolanda has joined #openstack-shade19:51
*** abregman|afk is now known as abregman21:43
*** mgagne_ is now known as mgagne23:12

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!