Thursday, 2016-01-14

« Wednesday, 2016-01-13 Index Friday, 2016-01-15 »
*** Qiming has joined #openstack-zaqar01:20
*** jasondotstar has quit IRC01:24
*** jasondotstar has joined #openstack-zaqar01:29
*** stanchan has quit IRC01:54
Eva-iflwang1: flwang:01:55
Eva-iping01:55
flwang1Eva-i: yes?01:56
Eva-iwhy Zaqar hasn't pymongo in requirements.txt and only in test-requirements.txt?01:56
Eva-i*and has only in test-requirements.txt?01:57
flwang1since it's a storage related lib01:57
flwang1user may use sql + redis01:58
Eva-iflwang1: okay, I understand01:58
flwang1:)02:00
*** Qiming_ has joined #openstack-zaqar02:01
Eva-iflwang1: I would include this library though, it's is very small and removes obligation from the user to install it manually.02:02
flwang1Eva-i: i'm reluctant to consider it in Mitaka :)02:03
flwang1seems therve's patch works for the subscriptions functional test issue02:03
flwang1but we're running into the keystone middleware issue again :(02:03
Eva-iflwang1: yes, I tested it and now figure out what exactly he did and what side-effects it might have02:04
Eva-iflwang1: yes, I noticed keystone issue02:04
*** Qiming has quit IRC02:04
*** AAzza has quit IRC02:05
flwang1Eva-i: what's the side-effects?02:08
Eva-iflwang1: just finished reviewing, no side effects found.02:09
*** AAzza has joined #openstack-zaqar02:09
flwang1Eva-i: hah, cool02:10
Eva-iflwang1: but I think I need to talk with therve about this patch02:10
flwang1ok02:10
Eva-iflwang1: I'll try to find solution to keystone problem now02:16
*** boris-42 has joined #openstack-zaqar02:18
flwang1Eva-i: cool, me too02:20
flwang1i just had a discussion with keystone guys, it's caused by this patch  https://review.openstack.org/#/c/253972/02:20
openstackgerritThai Tran proposed openstack/zaqar-ui: Zaqar-ui base project with empty queues panel  https://review.openstack.org/25521302:36
Eva-iflwang1: did they proposed any solution?02:37
flwang1Eva-i: no, keystone guys are working on that02:37
Eva-iflwang1: so it's only their problem?02:37
flwang1Eva-i: they said there is a solution02:37
flwang1but i haven't see it yet02:38
flwang1Eva-i: both?02:38
flwang1it's related the way how zaqar use config02:38
flwang1you now, we don't have a global config object02:38
Eva-iI thought we use global cfg.CONF: zaqar/cmd/server.py:28, but then just pass this object manually through hierarchy, pretending this is not grobal object.02:43
*** davideag_ has joined #openstack-zaqar02:51
*** davideagnello has quit IRC02:52
*** mdnadeem has joined #openstack-zaqar02:56
Eva-iflwang1: Should we wait now until keystone team will make changes on their side?02:59
Eva-iflwang1: or we must cooperate?02:59
flwang1Eva-i: here is the fix https://review.openstack.org/#/c/267277/103:03
openstackgerritFei Long Wang proposed openstack/zaqar: Fix redis subscription loading  https://review.openstack.org/26683103:05
flwang1Eva-i: i just uploaded a new patchset to depend on the keystone fix, if it can pass, then all good03:06
Eva-iflwang1: wow, nice idea!03:07
Eva-iflwang1: Depends-On is pretty convenient03:11
Eva-iflwang1: I'm going to sleep now03:12
Eva-igood bye03:12
flwang1Eva-i: good night03:13
openstackgerritFei Long Wang proposed openstack/zaqar-ui: Zaqar-ui base project with empty queues panel  https://review.openstack.org/25521303:34
openstackgerritFei Long Wang proposed openstack/python-zaqarclient: Support claims for v2  https://review.openstack.org/26566303:40
openstackgerritFei Long Wang proposed openstack/python-zaqarclient: Expose more CLI commands for v2  https://review.openstack.org/25875003:42
*** Qiming_ has quit IRC03:54
*** flwang has quit IRC04:07
*** Qiming_ has joined #openstack-zaqar04:11
*** stanchan has joined #openstack-zaqar04:44
*** stanchan has quit IRC04:52
*** kgriffs is now known as kgriffs|afk05:30
*** kgriffs|afk is now known as kgriffs05:31
*** kgriffs is now known as kgriffs|afk05:59
*** ig0r_ has joined #openstack-zaqar09:17
*** ig0r_ has quit IRC09:41
*** Qiming_ has quit IRC10:39
*** GB21 has joined #openstack-zaqar11:05
openstackgerritThomas Herve proposed openstack/zaqar: Fix redis subscription loading  https://review.openstack.org/26683111:08
*** GB21 has quit IRC11:15
*** Qiming_ has joined #openstack-zaqar11:30
*** Qiming_ has quit IRC11:30
*** Qiming_ has joined #openstack-zaqar11:30
*** Qiming_ has quit IRC12:14
*** Qiming_ has joined #openstack-zaqar12:14
openstackgerritHarshada Mangesh Kakad proposed openstack/python-zaqarclient: Replace deprecated library function os.popen() with subprocess  https://review.openstack.org/26754213:06
*** malini has joined #openstack-zaqar13:43
*** amitgandhinz has joined #openstack-zaqar13:47
*** dynarro has joined #openstack-zaqar14:15
*** wanghao has quit IRC14:17
*** njohnston has joined #openstack-zaqar14:42
*** tej has joined #openstack-zaqar14:42
*** tej_ has joined #openstack-zaqar14:42
*** mpanetta has joined #openstack-zaqar14:43
*** GB21 has joined #openstack-zaqar14:46
*** tej_ has quit IRC14:50
*** tej has quit IRC14:50
*** pradk has joined #openstack-zaqar15:11
*** Eva-i has quit IRC15:17
*** Eva-i has joined #openstack-zaqar15:19
*** kgriffs|afk is now known as kgriffs15:23
*** amitgandhinz has quit IRC15:35
*** Qiming_ has quit IRC15:38
*** amitgandhinz has joined #openstack-zaqar15:46
*** amitgandhinz has quit IRC15:55
*** amitgandhinz has joined #openstack-zaqar15:57
*** GB21 has quit IRC16:00
*** mdnadeem has quit IRC16:35
*** csoukup has joined #openstack-zaqar16:48
*** dynarro has quit IRC17:06
openstackgerritThomas Herve proposed openstack/zaqar: Fix gate issues  https://review.openstack.org/26683117:31
*** flwang has joined #openstack-zaqar17:48
*** malini has quit IRC17:49
flwangtherve: ping17:52
flwangre the gate issue17:52
*** AAzza has quit IRC17:59
*** GB21 has joined #openstack-zaqar17:59
*** AAzza has joined #openstack-zaqar18:01
*** flwang has quit IRC18:05
*** kgriffs is now known as kgriffs|afk18:05
therveflwang1, Hi18:26
*** malini has joined #openstack-zaqar18:28
*** malini1 has joined #openstack-zaqar18:30
*** malini has quit IRC18:32
Eva-itherve: ping18:53
Eva-iflwang1: ping18:54
therveEva-i, Hey what's up18:59
therveEva-i, https://blogs.gnome.org/markmc/2014/02/20/naked-pings/ future reads :)19:01
Eva-itherve: hello, in gate fixing patch you added this line to uwsgi.conf: add-header "Connection: close". How did you realised this will solve the issue?19:01
therveMostly, superpowers19:02
Eva-i:o19:02
therveHeh19:02
therveEva-i, So, I managed to track down the issue to HTTP pipelining support19:03
therveBecause subscription create reuses the HTTP session19:03
therveSee here: https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L6019:03
therveUsing some googling, I noticed that uwsgi didn't support it very well19:04
openstackgerritThai Tran proposed openstack/zaqar-ui: Zaqar-ui base project with empty queues panel  https://review.openstack.org/25521319:04
therveAt first, I tried to make it work by settting http-keepalive to true, and add the appropriate header on the response19:04
therveBut it didn't work19:04
therveSo instead I thought about how to make it so that it doesn't use pipeling19:05
therveAnd adding connection close is a way to tell the client: "please reopen the connection", instead of reusing it19:05
therveEva-i, Does that make sense?19:05
Eva-itherve: https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L60 this line doesn't tell me that the client will reuse the connection19:06
therveEva-i, See the trans and req variables19:07
therveThey are previously used here: https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L4919:07
therveI also used tcpdump to confirm what I saw19:08
therveie uwsgi closed the connection without a response19:08
Eva-itherve: yes, now it makes sense19:09
Eva-itherve: so now with "Connection: close" header from Zaqar, the clients are forced to stop http session after each request and don't use pipelining?19:11
therveEva-i, That's correct19:11
Eva-itherve: respect man19:12
Eva-itherve: you're the real programmer19:12
therveshrugs19:12
therveEva-i, I just looked at a bit more lines of code than you did19:12
therveThat's what age gets you :)19:13
*** malini1 has quit IRC19:22
Eva-itherve: thank you about the article about naked pings. I will not use them anymore.19:23
Eva-i*thank you for the article19:24
therveNo worries19:30
*** malini has joined #openstack-zaqar19:31
*** kgriffs|afk is now known as kgriffs19:36
*** njohnston has quit IRC19:38
*** njohnston has joined #openstack-zaqar19:41
*** amitgandhinz has quit IRC19:43
*** amitgandhinz has joined #openstack-zaqar19:43
*** malini has quit IRC19:44
*** flwang has joined #openstack-zaqar19:48
openstackgerritMerged openstack/zaqar-ui: Zaqar-ui base project with empty queues panel  https://review.openstack.org/25521319:57
flwang1yeah!!! the first commit of zaqar-ui merged :D19:57
Eva-iflwang1: congratulations =)19:58
Eva-iflwang1: why jenkins do not check this patch? https://review.openstack.org/#/c/26683119:59
Eva-i*does not19:59
flwang1Eva-i: i did, just because it's still in progressing20:01
flwang1and i think therve's patch works20:01
*** malini has joined #openstack-zaqar20:01
Eva-iflwang1: mmm, I don't understand "still progressing". I thought jenkins should check the patch on each patch set upload. Something blocks it?20:04
flwang1http://status.openstack.org/zuul/20:05
flwang1search zaqar20:05
Eva-iflwang1: I understand now, thank you.20:07
flwang1:)20:07
flwang1therve: still around?20:17
therveflwang1, Yep!20:18
flwang1as for the gate fix, i'm wondering if user will run into it if they use uwsgi + zaqar20:19
therveYeah, probably20:19
Eva-iflwang1: maybe to completely eliminate this potential problem for user, we should make Zaqar add connection close header20:20
flwang1therve: so maybe i should fix it in zaqar client20:20
therveflwang1, The bug is in the client though20:20
therveUrg20:20
therveflwang1, The bug is in uwsgi20:20
flwang1based on your investigation, i think we could just create a new connection for subscription get instead of reuse the one for create20:21
flwang1given subscription actions is not a performance matter action like message's20:21
flwang1so it should be fine, thoughts?20:21
therveWhat we're trying to do is perfectly fine20:22
flwang1therve: yep, i see20:22
flwang1i mean to avoid this issue, we can just get the req, trans again before https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L6020:23
flwang1does that make any sense?20:23
Eva-iflwang1: yes, it makes20:23
therveI understand, but I don't see a reason to workaround a bug with a specific configuration of a specific deployment20:24
*** flwang has quit IRC20:24
therveIf you use uwsgi and zaqar just use our tested configuration :)20:25
flwang1therve: haha20:25
flwang1ok, you sold it20:25
flwang1but i will document it somewhere :)20:25
flwang1in case we forget it20:26
therveSeems reasonable yeah20:26
flwang1btw, does the configuration impact any performance?20:27
*** flwang has joined #openstack-zaqar20:27
therveSure20:27
*** GB21 has quit IRC20:28
Eva-idoes zaqar now really allow reusable http connections, i.e. when zaqar runs without uwsgi and we create subscription in the client, these two requests from zaqar are going through a single connection?20:28
therveProbably not by much though. I guess it'd be nice to document a better wsgi container20:29
flwang1Eva-i: i don't think it's related to zaqar based on current investigation20:29
therveEva-i, Good question.20:30
therveEva-i, I'd be surprised if wsgiref supports http pipelining20:30
Eva-iflwang1: I'm just asking if Zaqar have http pipelining feature20:30
therveIt doesn't really need to, its a job for the container (AFAIK)20:31
flwang1hmm....20:32
flwang1IIRC http Post can't be pipling only the idempotent method can do that20:33
flwang1our subscription create using POST20:33
flwang1and subscription get is HTTP GET20:33
flwang1so maybe we have to change the client code?20:33
therveI guess I rather mean persistent connections20:34
thervePipelining can be tricky with POST indeed20:35
therveThe client actually doesn't do pipelining here, it uses a persistent connection20:35
therve(My bad for not using the proper term)20:35
flwang1yep, persistent connection20:36
Eva-iin uwsgi this is called "http-keepalive"20:38
Eva-i"HTTP Keep-Alive"20:38
therveAnyhow, patch is green, so hooray20:38
Eva-iyay20:38
flwang1therve: i will give it a manual try before +2 :)20:40
therveflwang1, Sweet!20:40
* therve away20:40
flwang1therve: thanks a lot20:40
flwang1Eva-i: can you try again for this one https://review.openstack.org/258750 ?20:46
Eva-iflwang1: oki, I will20:46
flwang1Eva-i: thank you very much20:46
flwang1since you said it doesn't work on your local20:47
Eva-iflwang1: how do you think, would it be good if we substitute this line https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L59 to "if self.id and not self.subscriber"? And also throw an error in these cases: 1. both id and subscriber are passed 2. both id and subscriber are not passed.20:56
flwang1Eva-i: wow, happy to see you notice this. I'm going to upload a new patch set to fix it20:58
flwang1and at line https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v2/subscription.py#L3220:59
Eva-iflwang1: I'm sorry I haven't noticed it during review of subscription create patch20:59
flwang1when there is a 'id' passed, the auto_create shouldn't be true20:59
flwang1or it will be skipped20:59
flwang1Eva-i: it's not a big problem, i would say21:00
flwang1but it should be improved21:00
flwang1thanks for the heads up21:00
Eva-iflwang1: ;)21:00
flwang1and i'm sorry to push you guys to review it in rush21:00
flwang1since i won't miss the integration with ceilometer in Mitaka21:01
Eva-iflwang1: you do things right as team leader21:01
flwang1Eva-i: thank you, and i believe i'm doing evil, you guys will poke me for sure21:02
flwang1missed 'if'21:03
flwang1Eva-i: so you tried the patch and it works, right?21:17
flwang1i saw you +1 on that21:17
Eva-iflwang1: yes21:17
flwang1cool, cheers21:17
Eva-iflwang1: hm...21:19
flwang1what?21:19
Eva-iflwang1: please wait until merging it21:19
flwang1no problem21:20
openstackgerritThai Tran proposed openstack/zaqar-ui: Re-enabling checkout and small fixes  https://review.openstack.org/26782021:21
*** njohnston has quit IRC21:28
*** mdnadeem has joined #openstack-zaqar21:30
*** mdnadeem has quit IRC21:38
*** shakamunyi has quit IRC21:40
*** njohnston has joined #openstack-zaqar22:15
*** mpanetta has quit IRC22:20
*** amitgandhinz has quit IRC22:40
Eva-iflwang1: "so you tried the patch and it works, right?" I should have respond "no" to this question, because I just tested that "openstack help --os-queues-api-version=2" do not fail anymore. Now I reviewed this patch once properly and found bugs.22:40
Eva-iflwang1: https://review.openstack.org/258750 maybe the patch has earned +1, because it's purpose is just to add two CLI commands. It's not the fault of this patch that the underlying things in CLI commands do not work well.22:45
flwang1ok, i see. so what's the problem?22:50
Eva-iflwang1: see my last review for this patch for problems22:50
flwang1"openstack flavor list" it's wrong22:54
flwang1you should use openstack messaging flavor list22:54
Eva-iflwang1: oh, you're right22:55
flwang1Eva-i: when you run openstack queue stats my-lovely-queue, you didn't specify the api version, right?23:00
Eva-iI tried it without specifying api version and tried it with all three api versions. The outcome is the same in all cases.23:01
flwang1ok, interesting...23:02
flwang1that's a good catch23:02
*** pradk has quit IRC23:04
flwang1Eva-i: i found the root cause23:06
Eva-iflwang1: should this patch be marked as +1?23:06
Eva-iflwang1: yes, what it is?23:07
flwang1the issue it's not related to the patch, i think23:07
flwang1Eva-i:  see line 205 of v1/cli.py23:09
*** njohnston has quit IRC23:13
Eva-iflwang1: for some reason self.client.api_version is not 1 when we specify it by "--os-queues-api-version=1": https://github.com/openstack/python-zaqarclient/blob/master/zaqarclient/queues/v1/queues.py#L5923:14
*** pradk_ has joined #openstack-zaqar23:15
flwang1Eva-i: really?23:16
flwang1i can't reproduce23:16
Eva-ihm wait23:17
Eva-iI'm wrong again...23:18
flwang1Eva-i: i will fix the stats issue with a separate patch. but i think queue stats works fine if you specify api version with 123:23
Eva-iflwang1: self.client.api_version is 1. But it has string type, so maybe the problem is in it. When I try this condition, the problem goes away:23:24
Eva-iif float(self.client.api_version) >= 1.123:24
flwang1haha, good catch23:24
*** stanchan has joined #openstack-zaqar23:25
Eva-iflwang1: strange that you can't reproduce it. When I try api version 1, I get this: http://pastebin.com/FHzimaYL23:26
Eva-iflwang1: should I open a bug?23:27
flwang1i can see '1'23:28
flwang1so pls file a bug23:28
flwang1and i'm happy to fix it, thanks23:28
Eva-iI'd be happy too, but oki23:29
Eva-iit's so easy to fix =)23:29
*** malini has quit IRC23:39
*** pradk_ has quit IRC23:40
*** Qiming has joined #openstack-zaqar23:46
*** flwang_home has joined #openstack-zaqar23:47
Eva-iflwang1: when I post a new bug report, should I put long logs to paste.openstack.org?23:53
Eva-iflwang1: here's the promised bug report: https://bugs.launchpad.net/zaqar/+bug/153437823:53
openstackLaunchpad bug 1534378 in zaqar "If client api version is passed string, some comparisons do not work properly" [Undecided,New]23:53
*** Qiming has quit IRC23:58
*** csoukup has quit IRC23:59
« Wednesday, 2016-01-13 Index Friday, 2016-01-15 »

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