Monday, 2017-04-10

*** tovin07_ has joined #openstack-oslo00:43
*** kiseok7 has joined #openstack-oslo00:53
*** nicolasbock has joined #openstack-oslo01:10
*** zhangguoqing has joined #openstack-oslo01:59
*** yamamoto has joined #openstack-oslo02:45
*** nicolasbock has quit IRC02:53
*** yamamoto has quit IRC03:01
*** yamahata has joined #openstack-oslo03:21
*** ctp has quit IRC03:45
*** ctp has joined #openstack-oslo03:46
*** zhangguoqing has quit IRC04:00
*** zhangguoqing has joined #openstack-oslo04:00
*** yamamoto has joined #openstack-oslo04:01
*** zhangguoqing has quit IRC04:05
openstackgerritMerged openstack/oslo.config master: Make set_override/set_default with enforce_type=True by default  https://review.openstack.org/32869204:19
*** Dinesh_Bhor has joined #openstack-oslo04:22
*** zhangguoqing has joined #openstack-oslo04:31
*** zhangguoqing has quit IRC04:32
*** zhangguoqing has joined #openstack-oslo04:32
*** ctp has quit IRC04:36
*** ctp has joined #openstack-oslo04:36
*** dharinic has quit IRC04:54
*** kornica is now known as kornicameister04:59
*** bknudson_ has quit IRC05:01
*** rcernin has joined #openstack-oslo05:25
*** yamamoto_ has joined #openstack-oslo05:42
*** yamamoto has quit IRC05:45
*** gcb has joined #openstack-oslo05:57
*** pcaruana has joined #openstack-oslo06:01
*** hieulq has quit IRC06:07
*** amotoki has joined #openstack-oslo06:23
*** tesseract has joined #openstack-oslo06:48
*** jaosorior has joined #openstack-oslo06:49
*** dmellado has joined #openstack-oslo07:08
openstackgerritMerged openstack/oslo.log master: protect systemd class initialization when syslog is not available  https://review.openstack.org/45427107:13
openstackgerritMerged openstack/oslo.log master: Always create OSSysLogHandler  https://review.openstack.org/45466407:13
*** dimtruck is now known as zz_dimtruck07:18
*** shardy has joined #openstack-oslo07:36
*** gcb has quit IRC07:46
*** lpetrut has joined #openstack-oslo07:48
*** zzzeek has quit IRC08:00
*** zzzeek has joined #openstack-oslo08:01
*** hieulq has joined #openstack-oslo08:06
*** gcb has joined #openstack-oslo08:12
*** sambetts|afk is now known as sambetts09:29
therveEmilienM, You still around? I had one question about https://review.openstack.org/#/c/454897/09:43
openstackgerritChangBo Guo(gcb) proposed openstack/oslo.log master: add an extras dependency for systemd  https://review.openstack.org/45427209:44
*** tovin07_ has quit IRC09:47
*** tovin07 has quit IRC09:48
*** tovin07_ has joined #openstack-oslo09:56
*** aarefiev_afk is now known as aarefiev09:59
*** tovin07_ has quit IRC10:04
*** boden has joined #openstack-oslo10:05
*** sdague has joined #openstack-oslo10:10
openstackgerritMerged openstack/oslo.log master: Documentation for journal usage  https://review.openstack.org/45414510:13
openstackgerritMerged openstack/oslo.i18n master: Updated from global requirements  https://review.openstack.org/45451710:15
*** gcb has quit IRC10:16
openstackgerritMerged openstack/oslo.messaging master: [zmq] Prevent access to rpc_response_timeout  https://review.openstack.org/44875310:17
*** hieulq has quit IRC10:19
*** nicolasbock has joined #openstack-oslo10:22
*** hieulq has joined #openstack-oslo10:33
*** sdague_ has joined #openstack-oslo11:04
*** shardy is now known as shardy_lunch11:06
*** cdent has joined #openstack-oslo11:06
*** lucasagomes is now known as lucas-hungry11:15
*** jamielennox|away is now known as jamielennox11:25
*** mtanino has joined #openstack-oslo11:51
*** shardy_lunch is now known as shardy12:07
*** lucas-hungry is now known as lucasagomes12:12
openstackgerritMerged openstack/oslo.log master: add an extras dependency for systemd  https://review.openstack.org/45427212:23
*** kgiusti has joined #openstack-oslo12:39
*** cdent_ has joined #openstack-oslo12:45
*** cdent has quit IRC12:46
*** cdent_ is now known as cdent12:46
*** gcb has joined #openstack-oslo12:50
*** ansmith has joined #openstack-oslo12:59
EmilienMtherve: hey! what's up?13:04
therveEmilienM, So there is one question I think I asked in the ML13:04
therveEmilienM, WRT to using etcd, won't it be confusing when using httpd?13:04
therveSome stuff will be configured by files, some by etcd13:05
therveFor the same "service"13:05
EmilienMtherve: I don't know how to tell the wsgi service to pull config from etcd13:07
EmilienMtherve: I don't even know if it's possible, tbh13:07
EmilienMbut for the vhosts, we will have to use config files like before, and the config files can be managed by confd13:07
therveEmilienM, I mean, if the wsgi service uses oslo.config, it will pull the config from etcd13:09
therveBut not the vhost config, obviously13:09
EmilienMtherve: right but how the wsgi process is run with the etcd url?13:09
EmilienMtherve: regarding the "confusion" - do you think it's something critical?13:12
therveEmilienM, Oh, that's an excellent point...13:12
therveWe can't pass the etcd url when running under WSGI13:12
EmilienMmaybe we can but I don't know how13:12
therveWhich means we need to create the config file for the service13:13
EmilienMand put the etcd url in a section?13:13
therveI *guess*13:13
EmilienMwe need to mention that in the spec13:13
therveBut yeah I think it's critical, even more after chatting with you :)13:13
*** eck`gone is now known as eck`13:14
EmilienMtherve: I agree it might be confusing but I still believe in the benefit of this feature. Think about it again, we don't need Puppet / Ansible / Chef (or whatever layer) to apply the configuration (and all the thousands parameters)13:15
therveEmilienM, Sure, I'm just making sure we don't miss something obvious13:16
therveThe point is that confd *needs* to be in the equation13:16
EmilienMfor sure13:17
EmilienMwell in fact no13:17
therveIf it can be the only solution, it may have some benefits, instead of having to deal with etcd inside the process themselves13:17
EmilienMyou don't need confd actually13:17
EmilienMlike if :13:17
EmilienM1) I deploy Keystone using httpd and managing the config file with Ansible or Puppet ...13:17
EmilienM2) I deploy Nova services (api not in wsgi, not officially supported yet) where the config is in etcd13:18
EmilienMin that scenario, we don't need confd in the equation13:18
therveWell sure if you keep puppet13:19
therveThat's even worse though, you know have 2 sources of truth13:20
therve2) is not a scenario. You will have services using httpd13:20
EmilienMtherve: which ones?13:20
*** zz_dimtruck is now known as dimtruck13:22
EmilienMthe thing with this spec is that 1) it's optional and disabled by default and 2) it's flexible enough to let deployers use it for some services at their wish13:22
EmilienMin tripleo, we'll probably start using it for the non wsgi services13:22
EmilienM(deployed in containers)13:22
EmilienMand we'll probably keep the configuration file at the beginning for the wsgi services, until we switch them to use confd (with etcd backend)13:23
therveYeah, but the end goal doesn't feel fantastic13:23
EmilienMsince apache needs a config file, we still need to find a way to manage it13:23
therveWe add code complexity without deployment improvements13:23
EmilienMsome end goals that I see :13:24
*** dtardivel has joined #openstack-oslo13:24
EmilienMcontainers running non-wsgi services don't need configuration file, and just their application13:24
EmilienMwhere each container would read the config from etcd and send logs into distributed logging system (with local logging of course but still sent to elk)13:25
EmilienMso we don't need puppet to generate this config (it takes time and we have 40 modules to maintain)13:25
EmilienMI think we'll save deployment / upgrade time for sure13:25
EmilienMthe second benefit I see on the long term is13:25
EmilienMdeployment projects (TripleO, OSA, Kolla, etc) would have a common and consistent way to make configuration management and won't maintain X or Y ways to do it (Puppet, Chef, Kolla Ansible templates, OSA templates, etc etc)13:26
EmilienMon the long term, we wouldn't need a tool to create config files13:27
therveEmilienM, That's the last part I'm questioning13:28
therveHow do you get rid of a tool to generate config files?13:28
EmilienMby removing config files?13:28
*** mtanino has quit IRC13:28
EmilienMlet me summarize 2 situations :13:29
EmilienM1) current situation, with tools like Puppet, Ansible, etc: managing configuration files13:30
EmilienMif you want to support a new option in TripleO / Kolla, etc, you need to patch the puppet module, the puppet-tripleo module (sometimes) and the THT (in kolla, you need to patch the config file template, which is horrible to maintain on longterm)13:31
EmilienMthe tool will compile a catalog and generate a config file13:31
EmilienMfor containers, you need to either put the config file in the container or mount it to the container so it's exposed13:31
EmilienM2) with the spec now: using etcd13:32
EmilienMoptions are generated with bnemec's spec and exposed to the deployment tool (tripleo, kolla, etc) via their interface (in tripleo it would be THT I guess)13:32
therveEmilienM, OK, but what about wsgi services that use httpd? You need config files for those?13:33
EmilienMthe deployment tool would write what they want/need in etcd for each project (ex: configure nova-scheduler)13:33
*** openstackgerrit has quit IRC13:33
EmilienMand the deployment tool would just have to start the containers, and the config is pulled from etcd13:33
therveI understand all of that :)13:33
EmilienMso you don't have to do anything everytime a new option is added to OpenStack13:33
EmilienM(in theory, but in fact some options might be computed and calculated by deployment tool)13:34
therveBut you're talking about a world without config files, and I don't see a way to get to that13:35
EmilienMin 2), we can see that Kolla / TripleO / OSA, etc would share 2 things here, that are not shared today: the parameters collection and the configuration reading from the non-wsgi services13:35
*** wuyanjun has quit IRC13:35
EmilienMno no no we still need config files wait13:35
EmilienMI just say, *some* services wouldn't need it13:35
*** wuyanjun has joined #openstack-oslo13:35
EmilienMfor example, most of non-wsgi services in OpenStack wouldn't need it13:36
EmilienMfor wsgi services, we still need a tool to maintain the httpd/nginx templates13:36
therveYou said "on the long term, we wouldn't need a tool to create config files"13:36
EmilienMand this tool could be a common project in OpenStack13:36
therveAnd I'm saying that's not the case, we would always need such a tool13:36
EmilienMsorry I was wrong13:36
therveOK :)13:37
EmilienMwe could imaging a new project in OpenStack to manage OpenStack WSGI files for httpd or nginx in confd13:37
therveSo my point is that I'm not sure there is great benefit to get into a hybrid situation13:37
EmilienMby providing a set of common templates, that can be re-used by different deployment tools13:37
EmilienMwell, I see different benefits :13:38
EmilienM1) allow to migrate containers (kubernetes) without dealing with config files13:38
EmilienM2) stop storing passwords and URLs in config files (security)13:38
EmilienM3) stop spending time at configuring files with puppet / ansible (it's not huge but still a win)13:39
*** dimtruck is now known as zz_dimtruck13:39
EmilienM4) expose a shareable method for config management to all deployment tools in OpenStack13:39
EmilienMthat's all I've been thinking now13:40
therveYou keep rehashing that13:40
therveAnd I keep repeating: you haven't manage to get rid of config files13:41
therveJust partially for some services13:41
EmilienMI should restate :13:41
EmilienM1) allow to migrate containers (kubernetes) without dealing with config files for most of OpenStack services13:41
therveBut is "most" good enough, that's the question13:42
EmilienMtbh, I don't know if this spec will solve the world but I think it's worth trying -- the current state of things is not so well13:43
EmilienMwe have a ton of duplication efforts around configuration management, around 40 modules to maintain, and dozen of thousands of Puppet code to keep in good shape13:43
therveYeah, and you will still those after the spec13:45
therveExcept for nova maybe13:45
EmilienMtherve: why wouldn't need the puppet modules to manage openstack configs?13:46
EmilienMtherve: except for wsgi files, that - like I said - could be managed by confd13:46
therveEmilienM, Because WSGI services load their own configuration too?13:46
therveEmilienM, Or you assume that the etcd url in the config files is already a solution13:47
EmilienMtherve: to me, it's something to investigate (the etcd url in config)13:47
therveYeah13:48
EmilienMand we wouldn't need puppet to manage this file, we can do it with something much light13:48
therveFair13:48
EmilienMI'm happy we had this discussion13:48
EmilienMwe have a bunch of challenges here13:48
therveYep me too :)13:49
therveI'll comment on the spec for that important detail13:49
EmilienMtherve: yeah, we need to collect this discussion output13:49
EmilienM1) this spec doesn't address OpenStack WSGI services, and we need a way to put the url of etcd in a config file13:50
EmilienM2) We need to mention that this spec is only useful for some services in OpenStack (the non-WSGI ones)13:50
*** Elaine_wu has joined #openstack-oslo13:50
EmilienM3) I don't know if we need to mention but we could write that for non-wsgi services, we could think at using confd etc (but to me it's out of the scope for the spec, since by design apache can't talk to etcd)13:51
EmilienMtherve: how does it sound ?13:51
*** sdague has quit IRC13:52
therveEmilienM, Those points seem to contradict one another13:52
*** wuyanjun has quit IRC13:52
therveIf you have the etcd url in a config file, that works for wsgi files13:53
therveIf you have the etcd url in a config file, that works for wsgi services13:53
*** ctp has quit IRC13:53
*** ctp has joined #openstack-oslo13:54
EmilienMtherve: OpenStack non-wsgi process would be run with url directly and OpenStack wsgi process could be run with classic /etc/keystone/keystone.conf with etcd_url parameter13:54
gcbweekly meeting in 5 minutes at channel openstack-meeting-313:55
*** sdague has joined #openstack-oslo13:56
*** sdague__ has joined #openstack-oslo13:56
EmilienMtherve: I'm about to board in the plane13:57
EmilienMtherve: thanks for this nice discussion13:57
EmilienMtherve: I'll catch-up on the spec review asap13:57
therveEmilienM, Have a good flight!13:58
EmilienMthanks!13:58
*** openstackgerrit has joined #openstack-oslo13:59
openstackgerritAndy Smith proposed openstack/oslo.messaging master: Add get_rpc_transport call  https://review.openstack.org/45419413:59
*** dave-mccowan has joined #openstack-oslo14:04
*** lhx_ has joined #openstack-oslo14:08
*** dave-mccowan has quit IRC14:09
*** sdague has quit IRC14:12
*** sdague__ is now known as sdague14:13
*** gordc has joined #openstack-oslo14:15
*** zz_dimtruck is now known as dimtruck14:15
*** mtanino has joined #openstack-oslo14:16
*** ihrachys has joined #openstack-oslo14:27
*** ihrachys has quit IRC14:30
*** ihrachys has joined #openstack-oslo14:30
ansmiththerve: hello14:39
*** zhangguo_ has joined #openstack-oslo14:42
*** zhangguoqing has quit IRC14:46
openstackgerritMerged openstack/devstack-plugin-amqp1 master: Use the Qpid released repo, not testing  https://review.openstack.org/45232014:46
therveansmith, Hi!14:47
*** zhangguo_ has quit IRC14:47
therveActually working on that amqp issue14:47
ansmiththerve: https://review.openstack.org/452793/14:48
therveansmith, Hummm tell me more? :)14:49
ansmiththerve: cmw was using the incorrect transport for notifications leading to timeouts that might be the root issue14:49
ansmiththerve: the timeouts to the driver were in the logs14:49
therveansmith, What does it look like?14:49
ansmiththerve, WARNING:oslo_messaging._drivers.amqp1_driver.controller:Notify message sent to <Target topic=notifications.info> failed: timed out14:51
ansmiththerve, as seen in syslog14:51
therveAh, syslog14:51
therveansmith, Does it need a release to be used in the gate?14:51
ansmiththerve, believe so14:52
therveOK cool14:52
ansmiththerve, have not asked for a cmw release14:52
therveansmith, I started looking at the results, and only a few tests are impacted14:52
therveNot sure if all are swift related, though14:52
*** erolg has joined #openstack-oslo14:55
erolghi everyone, i am try to run brocade vrouter neutron plugin. When I try to start, i am getting "no such option in group DEFAULT: nova_admin_auth_url" from oslo_config. I would say it is deprecated parameter but how can i reuse it?14:57
*** yamahata has quit IRC15:00
*** yamahata has joined #openstack-oslo15:00
*** rcernin has quit IRC15:03
thervegordc, Hi! Can we make a release to ceilometermiddleware to get the amqp fix?15:04
*** dharinic has joined #openstack-oslo15:10
*** aarefiev is now known as aarefiev_afk15:27
gordctherve: sure. i'll do it since jd is away15:30
*** jdandrea_ is now known as jdandrea15:38
*** chenying_ has quit IRC15:43
*** chenying_ has joined #openstack-oslo15:44
*** lpetrut has quit IRC15:47
*** lucasagomes is now known as lucas-afk16:05
*** jokke__ has joined #openstack-oslo16:11
*** jokke__ has quit IRC16:11
thervegordc, Thanks!16:19
*** tkill has joined #openstack-oslo16:21
*** lhx_ has quit IRC16:50
*** ctp has quit IRC16:51
*** jaosorior is now known as jaosorior_away16:51
*** tesseract has quit IRC16:52
*** ctp has joined #openstack-oslo16:53
*** chenying_ has quit IRC16:58
*** chenying_ has joined #openstack-oslo16:58
*** amotoki has quit IRC17:09
*** erolg has quit IRC17:20
harlowjajd_ any reservations putting cotyledon into G-R ?17:40
* harlowja convincing the octavia folks to use it17:40
harlowjabut noticed its not in G-R17:40
harlowjasileht ^17:40
johnsomharlowja What trouble are you stirring up now???  grin17:44
harlowjalol17:44
harlowjaidk17:44
harlowja:-P17:44
harlowjajohnsom freeing octavia from eventlet mainly :-P17:47
johnsomNo arguments there17:48
*** sambetts is now known as sambetts|afk18:00
silehtharlowja, you can add it18:03
harlowjakk18:03
silehtharlowja, telemetry does follow requirements repo, that why we didn't put it for now18:04
silehtdoes/does not18:04
harlowjaokie dokie, sileht https://review.openstack.org/45541118:08
*** dharinic has quit IRC18:10
*** kgiusti has left #openstack-oslo18:30
*** kgiusti has joined #openstack-oslo18:30
*** dharinic has joined #openstack-oslo18:50
*** sdague_ has quit IRC18:55
*** dtardivel has quit IRC19:00
*** edmondsw_ has joined #openstack-oslo19:15
*** edmondsw_ has quit IRC19:15
*** yamahata has quit IRC20:03
*** shardy has quit IRC20:03
*** dimtruck is now known as zz_dimtruck20:09
*** zz_dimtruck is now known as dimtruck20:31
*** kgiusti has left #openstack-oslo20:35
*** ansmith has quit IRC20:55
*** tkill has quit IRC20:57
*** boden has quit IRC21:12
*** pcaruana has quit IRC21:15
*** mortenover has joined #openstack-oslo21:44
*** tkill has joined #openstack-oslo21:55
*** hoonetorg has quit IRC22:16
*** gordc has quit IRC22:21
*** Elaine_wu has quit IRC22:23
*** Elaine_wu has joined #openstack-oslo22:23
*** ihrachys has quit IRC22:25
*** dimtruck is now known as zz_dimtruck22:28
*** zz_dimtruck is now known as dimtruck22:33
*** hoonetorg has joined #openstack-oslo22:39
*** mortenover has quit IRC22:48
*** sdague has quit IRC23:12
*** dharinic has quit IRC23:15
*** dimtruck is now known as zz_dimtruck23:20
*** zz_dimtruck is now known as dimtruck23:24
*** ansmith has joined #openstack-oslo23:26
*** cdent has quit IRC23:37
*** dharinic has joined #openstack-oslo23:40
*** dimtruck is now known as zz_dimtruck23:54
*** zz_dimtruck is now known as dimtruck23:54

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