19:03:29 #startmeeting Poppy Weekly Meeting 19:03:31 Meeting started Thu Aug 27 19:03:29 2015 UTC and is due to finish in 60 minutes. The chair is amitgandhinz. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:03:32 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 19:03:34 #topic Roll Call 19:03:35 The meeting name has been set to 'poppy_weekly_meeting' 19:04:19 o/ 19:04:21 o/ 19:04:30 o/ 19:04:38 o/ 19:04:41 #topic Last Week Recap 19:04:43 #link http://eavesdrop.openstack.org/meetings/poppy_weekly_meeting/2015/poppy_weekly_meeting.2015-08-20-19.03.html 19:04:59 malini to resurrect https://review.openstack.org/169611 19:05:08 look slike its still in conflict 19:05:22 we have another patch to supersede this one 19:05:31 does this need to be abandoned? 19:05:32 I'll kill it after the other is merged 19:05:43 where is the other one? 19:05:46 Or just track the new patch & I'll abandon this? 19:05:55 just track the new patch IMO 19:06:15 https://review.openstack.org/#/c/207988/ 19:06:55 go ahead an abandon it. i dont see a reason to wait 19:07:14 abandoned patches still live in a secret area 19:07:20 and can be reclaimed 19:08:05 malini to abandom 169611 after Henry's patch is merged 19:08:08 we just covered this 19:08:15 amitgandhinz to resurrect https://review.openstack.org/143192 19:08:22 #action amitgandhinz to resurrect https://review.openstack.org/143192 19:08:26 i have not done this 19:08:32 rolling it over 19:09:04 #topic Liberty 3 19:09:09 #link https://launchpad.net/poppy/+milestone/liberty-3 19:09:24 so one week left in the liberty 3 cycle. 19:09:39 https://blueprints.launchpad.net/poppy/+spec/gate-api-tests 19:09:47 this still has very slow progress, no updates on this 19:10:06 https://blueprints.launchpad.net/poppy/+spec/akamai-ssl-driver 19:10:08 tonytan4ever: ^ 19:10:30 This one is in good progress. The first one PR is ready for review 19:10:44 https://review.openstack.org/#/c/209648/ 19:11:17 It's a huge PR so we'd better start early and review it 19:11:31 yes everyone please review 19:11:33 Maybe I will start hounding people review it again. 19:11:58 please hound 19:12:02 tonytan4ever: https://blueprints.launchpad.net/poppy/+spec/geo-restrictions 19:12:14 This one is in review as well 19:12:29 I am actively working on it today, adding api tests there. 19:12:39 It should be ready by today. 19:12:43 cool 19:13:06 any thing else on bugs and bps? 19:13:44 #topic Open Discussion 19:14:06 so one of our talks has been accepted in Tokyo 19:14:08 yay! 19:14:13 yayyy.. 19:14:13 awesome 19:14:18 Woot! 19:14:25 Congrats 19:14:36 https://openstacksummitoctober2015tokyo.sched.org/event/d8c883ae776168eae3283f0b740b0027?iframe=yes&w=i:0;&sidebar=yes&bg=no#?iframe=yes&w=i:100;&sidebar=yes&bg=no 19:14:59 poppy is coming to Tokyo - be careful with the customes 19:15:17 ;) 19:15:53 tonytan4ever: do you want to bring up the issue with regards to purge_url hard and all? 19:16:01 its a shame its the last session of the summit though 19:16:02 Oh yes 19:16:11 hopefully we have stackers show up 19:19:23 what is the issue with the purgeurl/ 19:19:52 So we have an issue when we are doing cache invalidation on Akamai driver. 19:20:52 Our protocol of doing purge/cache invalidation is when we pass in purge_url with None, we treat it as purge/cache invalidate "ALL" 19:21:46 This brings a problem, when the user pass in a purge_url with a string "None" and doing cache invalidation, it will blow up. 19:22:04 why do we support "None"? 19:22:09 i dont see that in the docs 19:22:13 tonytan4ever: no, that is not the current functionality, but one of the suggested things right? 19:22:18 we have ?all=true 19:22:45 Yes, we do have all=true. 19:22:50 url=None does not seem valid 19:22:57 url=/* would be valid 19:23:11 or url=/images/cat.jpg 19:23:23 amitgandhinz: hard purging, looks at purge_url and if its find to be None, then raises a RunTimeError 19:23:45 I think url=None can be valid, a user could have /None as a url 19:23:58 tonytan4ever: yeah 19:24:11 if a user enters ?url=None, shouldnt poppy translate that into purging url=/None 19:24:18 where it thinks the url is /None 19:24:21 it does 19:24:32 does poppy throw the error, or akamai? 19:25:03 we threw the error, but thats been fixed in a follow up patch.. 19:25:16 which is in review right now. 19:25:43 I think what tonytan4ever is trying to point out, is that we transform "None" to None in case of hard purge.. 19:25:55 but not in the case of soft purge. 19:25:59 well ok 19:26:03 im confused. 19:26:16 i dont understand why 19:26:30 if the user passes a value of None,it should be a valid url. 19:26:37 ok what we have in the patch now works.. 19:26:37 what if i happen to have a file called "None" 19:26:46 and i want to purge it 19:26:53 amitgandhinz: that scenario works.. 19:27:33 we allow url to be set to "None" 19:28:12 this is setting the purge url to None though - https://review.openstack.org/#/c/217739/3/poppy/distributed_task/taskflow/task/purge_service_tasks.py 19:28:30 as opposed to purging a file called "/None" 19:28:41 That is a problem 19:29:18 We don't know if the parameter is passed as None object or "None" string 19:29:24 None and "None" should be different. one is reserved in python, one is a string 19:30:20 The root cause of this problem is that when we pass parameters in taskflow tasks, we cannot pass None object. 19:30:26 so if i enter url="None", will it purge everything or purge my file called "/None" 19:30:26 We can only pass string 19:31:15 neither, we will raise a runtimeError there. 19:31:20 and that is the bug. 19:31:25 ok, so when we set all=true, in that scenario, url=None, and taskflow changes that to "None", and then blows up? 19:31:26 Yeah 19:31:44 in case of cache invalidation -> everything works. 19:31:58 so why dont we set it to "/*" instead of "None" as the default? 19:32:24 that way we arent overloading the meaning on "None" 19:32:29 We need to confirm /* works with purge and cache invalidation. 19:32:34 that is one of the suggested solutions. 19:32:43 Which is not the case with purge. 19:32:45 but malini suggested that it doesnt work. 19:32:47 it wont work with purge as purge requires a file 19:33:04 but our validation should catch that 19:33:16 purge should always require a url path 19:33:24 invalidate can have the wildcards 19:33:43 But purge can have wild cards for other providers - eg. fastly 19:33:53 thx to sriram for tht info ^ 19:34:08 true 19:34:12 the issue is, if we enforce it at transport 19:34:26 it might seem like we are biased towards providers 19:34:39 ya 19:34:44 we shouldnt retrofit provider needs into transport. 19:35:05 so lets default it to "/*" and if the provider doesnt support the wildcard purge then it fails 19:35:24 today we do that with defaulting it to None right? 19:35:36 yeah 19:36:15 by defaulting the url = "/*" we remove the overloading of the word None/"None", we become more explicit, and we arent changing overall behavior 19:37:00 we would still need to have a case for None though.. 19:37:14 because the user can still give a url of None 19:37:17 tonytan4ever: ^ 19:37:34 in which case we pass a string to taskflow which is "/None" as we have to add the preceding slash 19:37:35 If a user give a url of None, we treat it as normal url right > 19:37:38 the url is now always a string 19:37:48 and never the None object 19:38:04 the code where we default url=None now becomes url="/*" so always is a string 19:38:35 Another thought: 19:38:50 if we default url to "/*", do we still need all though > 19:38:51 ? 19:38:53 yeah, its always a string, since we put it into zookeeper 19:38:56 all=true ? 19:39:09 tonytan4ever: again that depends on the provider 19:39:19 some provider will have an all option 19:39:23 some might not 19:40:07 I understand that part. But url /* means purge all right > 19:40:08 ? 19:40:17 it shoud 19:40:26 malini: ^^ 19:41:01 yes 19:41:41 all=true is technically redundant, but we have to support it now since its out there 19:42:50 we can inform our clients ;) 19:43:09 testing for purge_url as None inside purge_task will need to change then.. 19:43:22 if that changes, we should be able to fix the rest. 19:43:41 yup 19:44:01 so everyone agree to default to /* and get rid of None ? 19:44:08 +1 19:44:17 +2 19:44:34 +22,345 19:44:48 We will need to adjust akamai driver though since it does not support /* 19:44:56 raise an exception when url is /* 19:45:13 yeah we can do the same runtime error thing 19:45:20 Yup 19:45:22 ya, do whatever we do now if i say PURGE url="/images/*" 19:45:30 shouldnt really be a change 19:45:44 ok next topic 19:45:53 im done. 19:46:04 my head hurts 19:46:10 mine too 19:46:13 need water 19:46:16 I preserve mine 19:46:21 or poppy 19:46:36 I hear preservatives are bad. 19:46:48 chutneys are good tho 19:46:53 ? 19:47:19 Alcohol preserves well 19:47:28 ok I think we are derailing, anyone have anything else to discuss? 19:47:32 i think this is what malini really meant by preserving her head 19:47:33 http://s3.hipertextual.com/wp-content/blogs.dir/6/files/2011/07/futurama.s06e20.hdtv_.xvid-fqm.avi_000153361-e1311964381808.jpg 19:47:36 http://www.ebay.com/itm/281130301682?_trksid=p2060353.m2749.l2649&ssPageName=STRK%3AMEBIDX%3AIT 19:47:41 I actually bought this from ebay 19:48:02 im ending this meeting 19:48:04 its derailed 19:48:11 #nedmeeting 19:48:11 tht explains ur behavior tonytan4ever 19:48:11 well done tonytan4ever 19:48:14 ugh 19:48:17 lolol 19:48:19 #endmeeting