16:03:33 #startmeeting JOID weekly 16:03:33 Meeting started Wed Jul 6 16:03:33 2016 UTC. The chair is arturt. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:03:33 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:03:33 The meeting name has been set to 'joid_weekly' 16:03:43 #info Artur Tyloch 16:03:50 #info Dave Urschatz 16:03:53 #info Narinder Gupta 16:04:10 #link https://etherpad.opnfv.org/p/joid 16:04:15 #topic agenda 16:04:24 #info Stuart Mackie 16:07:32 #topic Colorad release status update 16:08:56 issues with 1.25 , fix expected in 1.25.6 16:15:34 #link https://jira.opnfv.org/browse/JOID-61 16:24:13 http://summit.juju.solutions/ 16:31:32 #topic OpNFV Summit Berlin and ODS plans 16:35:45 #topic JIRA Issues 16:37:49 #link https://jira.opnfv.org/browse/JOID-88 16:40:09 #link https://jira.opnfv.org/browse/JOID-80 16:41:14 #link https://jira.opnfv.org/browse/JOID-77 16:44:43 #link https://jira.opnfv.org/browse/JOID-65 16:52:10 #info we will skip next week JOID meeting 16:40:24 Hola 19:14:08 zahra_jahedi, lets do it as we type 19:14:21 lets change the power parameter of the bootstrap node 19:14:27 click on bootstrap node 19:51:28 jmorgan1, whenever you are back pod6 seems to be ok except creating gateway 10.6.15.254 19:51:59 jmorgan1, but node 6 in pod5 is not pxe booting from admin network. If you can have a look would be great? 14:26:22 durschatz, zahra_jahedi do we know whether cengen community lab is connected to jenkins or not? 16:03:59 narinder: I believe as of this morning the answer is YES. At least Faith was able to produce https://build.opnfv.org/ci/job/pod_test/95/console 16:04:42 Daniel who did the work on our side will not be back in the office till Monday. 16:11:26 durschatz, ok do we know which installer will be using this pod. 17:06:46 narinder: it will be joid 17:09:42 durschatz, ok if everything is ok then we can do a dry run today and how does it perform? 17:11:19 Is this dry run to help Zahra with her work? 17:12:47 The issue with the bootstrap KVM is happening on that POD 18:31:22 durschatz, we can try using it for C release itself as latest code is more stable specially with nonha mode 18:31:55 ha mode gets install but ssh into the unit through floating ip have few issues 18:33:33 narinder: Do you have time for a quick call? 19:55:54 narinder: We'll need to wait till Monday to discuss further. 19:57:48 sure no problem 13:25:56 #info rprakash 13:52:29 link https://gerrit.opnfv.org/gerrit/#/settings/ssh-key 01:57:57 narindergupta: ping 14:32:28 ci stuff will be voted during infra meeting 14:32:35 feel free to join 14:32:38 https://wiki.opnfv.org/display/INF/CI+Evolution#CIEvolution-WayForward 17:37:07 narindergupta: ping 15:18:47 narindergupta: ping 17:04:30 narindergupta: ping ping! 12:32:23 yuanyou_: hi 12:32:43 bryan_att: i was on vacation just rejoined today 12:32:58 fdegir: sorry i was on vacation last week 12:36:52 narindergupta: welcome back 12:37:02 narindergupta: it was about changes to CI-stuff 12:37:12 fdegir: thanks ok no problem 12:37:19 narindergupta: and I started doing updates to joid jobs 12:37:20 how did the vote went? 12:37:20 narindergupta: https://gerrit.opnfv.org/gerrit/#/c/17033/ 12:37:31 fdegir: yeah i saw the requesy 12:37:46 narindergupta: voting passed but it was just for reference purposes 12:38:02 fdegir: ok gotch you. 12:38:10 narindergupta: and I basically send changes to installer projects and if you/they accept, they get implemented 12:38:19 narindergupta: if not, then I can't do much 12:38:34 narindergupta: btw, as explained in the patch 12:38:40 narindergupta: nothing changes for joid as of yet 12:38:54 narindergupta: will ping you once we get the basics in place and before activating anything for real 12:39:08 fdegir: sure i would love to try 12:39:15 narindergupta: happy to heart 12:39:17 hear 12:39:22 fdegir: :) 12:39:56 narindergupta: please review/comment that patch whenever you finish more important stuff you might have in your mailbox 12:40:42 fdegir: yeah i have week backlog and working on clearing it one by one 12:41:16 good luck 12:54:18 hi all, I am trying to boot ubuntu VM using iPXE with MaaS without success. The error is "no bootable device" 12:55:55 does anyone already have this issue ? 13:02:19 virginie___: usually jois take care of creating a VM and adding into MAAS for bootstrap. And we are not observing such issue during deployment. how exactly you are using joid? 13:09:11 Maas created the bootstrap VM, but nothing has been installed 13:09:33 the Vm is blocked at the startup and said that "there is no bootable device" 15:03:12 virginie___: looks like you MAAS is not setup for pxe boot correctly. 17:13:34 bryan_att: bryan looks like we are finding issue with congress charm now i am facing this issue. congress/0 error idle 1.25.6 1/lxc/1 10.6.1.83 hook failed: "identity-service-relation-changed" for keystone:identity-service 17:13:51 bryan_att: can you have a look into it please? 17:16:07 narindergupta: I guess you were ooo last week. I am having issues with getting JOID installed. Persistent errors with the ceph charm is blocking install. I need your help to address that issue first, so I can get the system deployed to debug the congress issue. 17:16:56 bryan_att: yes i was OOO to cancun mexico 17:17:06 has nice vacation 17:18:05 today i back to work. yeah ceph charm issue is resolved now. I worked with openstack charmers team today and fix is available now at the charm. 17:18:46 narindergupta: thanks, I'll try again. 17:18:58 bryan_att: :) 17:20:27 bryan_att: looks like congress data base sync command is failing. http://paste.ubuntu.com/19922765/ 17:20:48 bryan_att: i can provide you my intel pod for debug if you are interested. 17:24:43 sure, send me the login credentials by email and I'll go there to debug 17:25:01 I have VPN access to the Intel lab 17:25:55 if you vpn access then ssh into 10.2.117.107 17:26:07 as jenkins user 18:13:57 narinder bryan_att: I had the same issue with JOID ceph charm. Since I need a quick NFVi I temp switched to a different installer. Please let us know when JOID is working again. 18:14:23 Also welcome back Narinder! 18:14:35 durschatz, currently ceph issue have been fixed asif afternoon 18:14:53 narindergupta: can you send me the jenkins account password? 18:14:54 currenly congress is broken and working with Bryan to fix it 18:15:07 Great! OK. 18:15:12 give me your public key i will add it to user account 18:15:32 i think i disabled the username / password access onyl key based in believe 18:15:52 bryan_att, if not try jenkins as password it seld\f 18:17:47 bryan_att, looks like automatic build started on that pod let me try pod5 18:18:29 narinder: ok, just found that out - "juju ssh congress/0 " doesn't work 18:20:05 bryan_att, beacsue automatic build started 18:20:17 wait for 30-45 minute you should see the error 19:37:33 bryan_att, pod5 i am seeing the same issue 19:37:49 bryan_att, you can ssh into it using 10.2.117.91 19:38:00 same user jenkins and same password 19:40:25 ok, I'll check it out 19:44:11 bryan_att, thanks 20:34:52 narinder: looks the database connection parameter in congress.conf did not get set correctly. Not sure why or how this gets set in the charm design but I will look into it. Is gnuoy available to help debug? https://www.irccloud.com/pastebin/OdbLSnRS/ 20:35:40 bryan_att, please check on #openstack channel someone or other should be available. 20:36:22 bryan_att, may be some api got changed in layer might impacting you 20:37:04 bryan_att, we can refer other charm how its done there but need to check which is connected to mysql 20:42:54 narinder: how do I know what charm was used to install congress? 20:44:09 bryan_att, i was looking into reactive charm for other charms it seems not there only one used is openvswitch-odl 20:44:47 but that charm do not have 20:44:49 code 20:45:22 in congress charm we have template and template value placed is shared.uri 20:45:31 but not getting how charm will get it 20:48:37 ongress/templates/parts/section-database has URL string 01:54:53 narindergupta: hi 02:52:18 yuanyou_: hi how r u? 02:52:53 narindergupta: hi 02:53:18 yuanyou_: can u tell me the changes neeeded in joid to integrete nsh 02:53:59 narindergupta: yeah, I had add profile named "onos-sfc" in charms openvswitch-onos,neutron-gateay 02:54:51 narindergupta: lp:~zhangyuanyou/onosfw/openvswitch-onos 02:54:57 yuanyou_: neutron-gateway why in neutron gateway? 02:55:04 narindergupta: lp:~zhangyuanyou/onosfw/neutron-gateway 02:55:32 narindergupta: because neutron-gateway installed openvswitch 02:55:57 yuanyou_: no neutron-gateway does not install openvswitch 02:56:14 yuanyou_: openvswitch-onos does 02:56:34 yuanyou_: but onos is not ovs dependent correct? 02:56:52 but i saw it has install-packages include "openvswitch" 02:57:25 that is only if neutron-openvswitch charm is installed 02:57:39 which is subordinate charm 02:58:34 the way it works is netutron-gateway will independent then we can integrate with any SDN 02:59:06 if gateway is provided by SDN then no need to deploy neutron-gateway charm 02:59:20 individual SDN gateway can be deployed 02:59:23 narindergupta: that's fine ,I can remove it from neutron-gateway 02:59:52 narindergupta: so I can only add onos-sfc to openvstich-onos 02:59:57 yuanyou_: i need to know what package you are replacing it with the new patches and which nodes 03:00:14 how this will get integreted with onos-controller? 03:00:49 yuanyou_: currently i am using the following charms and location bzr branch lp:~zhangyuanyou/onosfw/onos-controller $distro/onos-controller 03:00:49 bzr branch lp:~zhangyuanyou/onosfw/neutron-gateway $distro/neutron-gateway 03:00:50 bzr branch lp:~zhangyuanyou/onosfw/neutron-api-onos $distro/neutron-api-onos 03:00:50 bzr branch lp:~zhangyuanyou/onosfw/openvswitch-onos $distro/openvswitch-onos 03:01:54 narindergupta: i used openvswitch-onos integrated with onos-controller by set-manager 03:02:38 yuanyou_: ok so no chamged in onos-controller 03:02:43 narindergupta: i should install openvswitch 2.5.9 that included nsh-patch 03:03:09 narindergupta: yes,no changed in onos-controller 03:03:29 ok 03:03:57 yuanyou_: and how are you installing it? Are you building it or using PPA to install? 03:04:20 and which nodes need openvswitch 2.5.9 ? 03:04:22 narindergupta: i used dpkg 03:04:34 on trusty? 03:04:51 and it had issue with xenial correct. 03:05:00 narindergupta: compute-nodes and neutron-gateway node 03:05:12 so basically all nodes 03:05:19 narindergupta: yes,sfc should run in trusty 03:05:40 i was just wondering does onos uses openvswitch like odl? 03:05:44 narindergupta: yes, all nodes that ran ovs 03:06:39 narindergupta: there is some diff in sfc. 03:06:50 ok that make sense so profile in openswitch-onos will be installed on all nodes and will have relationship accordingly 03:08:18 yuanyou_: in current relationship openvswitch-onos is not connected to compute nodes only to neutron-gateway 03:09:07 yuanyou_: sorry my bad its there as plugin 03:10:01 narindergupta: i had test in my local env by the above charms. 03:10:03 yuanyou_: so i need to just enable the profile option in charm bundle correct and use trusty as default option for deployment this is correct. 03:10:50 can i send me yopur bundle you tested as that will give me clear picture what changes i need 03:11:30 narindergupta: ok, i will send you a email about i changed. 03:13:12 yuanyou_: you can do pastebinit also? 03:15:26 https://www.irccloud.com/pastebin/f5foWod7/ 03:16:04 narindergupta: this is i defined profile in openvswitch-onos charm 03:16:38 ok understood and in bundle just define this option correct 03:16:40 ? 03:17:04 narindergupta: that's right 03:17:13 ok 03:17:37 in \jenkins jous we need -d trusty 03:17:55 parameter for sfc with onos deployment then 03:19:18 narindergupta:then we should use sed -i -- 's/#profile: onos-sfc/profile: onos-sfc/g' ./bundles.yaml 03:19:36 yuanyou_: no thats not needed anymoe 03:19:39 anymore 03:19:53 we are creating the bundle dynamically now 03:20:10 change needed in bundles_tpl/subordinate.yaml 03:21:06 narindergupta: ok,another question? 03:21:13 yuanyou_: yeah 03:21:31 narindergupta: what's genBundle.py do? 03:22:07 narindergupta: it will gen a new bundles? 03:22:19 narindergupta: bundles.yaml 03:22:32 yuanyou_: that will create the bundles.yaml for deployment with juju-deployer during run time based on inputs providede 03:23:12 we have template for each charm and modify it based on feature requested etc... 03:23:22 narindergupta: but there is some problem such as some config option will disappear 03:23:37 yuanyou_: tell me which one i will fix it 03:24:03 yuanyou_: so far i did not find any issues but we will fix it 03:24:08 if it appears 03:25:09 narindergupta: onos-controller: 03:25:09 charm: local:trusty/onos-controller 03:25:09 options: 03:25:09 #ext-port: "eth1" 03:25:47 narindergupta: this options not set 03:25:49 i was not seeing any value in bundle as onos-controller is in container 03:26:07 and there is no eth1 on containers 03:26:35 but if it matches the neutron-gateway ext-port then i can fix it right away 03:27:19 narindergupta: yes 03:27:30 so you need it in onos-controller 03:27:52 narindergupta:yes, i need this for onos-controller 03:28:17 and the ext-port should be corrct ext-port 03:28:44 of neutron-gateway 03:28:52 as deifned in neutron-gateway 03:29:18 narindergupta: when is use -d trusty to deploy, and the bundles.yaml is still xenial 03:29:38 did u cleaned up? 03:29:45 before depoloyment? 03:30:12 i think i fixed this before going on vacation but will have a look again 03:30:21 narindergupta: ok 03:31:37 yuanyou_: do you need gateway-mac option also in onos-controller? 03:32:00 narindergupta: not need now 03:32:04 ok 03:32:36 i have fixed both ext-port and profile 03:32:43 any other option i am missing 03:33:10 narindergupta: good 03:33:29 yuanyou_: i will start the deplyment and send you bundles.yaml created on intel pod and once you okeyed it i will check in into joid 03:34:12 narindergupta: ok 03:38:02 yuanyou_: now bundle file will get created after the juju bootstrap during runtime 03:38:45 and reason we changed because earlier we used to make changes at 12 place for any change in charm 03:39:00 now only one place and integration is free of sed now 03:39:22 narindergupta: ok ,i will hava a look 03:39:30 yuanyou_: sure 03:39:44 all templates are in bundles_tpl 03:39:49 directory under ci 03:41:01 narindergupta: ok, i saw it 03:41:16 cool 03:42:30 i used the below command to depploy ./deploy.sh -s onos -d trusty -f sfc -t ha -l intelpod5 03:43:35 i am making few more changes where using few charm from charm store rather than git 03:43:49 but it should not affect your deploymen 03:48:04 narindergupta: ok ,and the ci time we should reset. 03:56:01 yuanyou_: give me 15 more minuutes for bundles.yaml as i made a mistake so retrying the fix now 03:57:59 narindergupta: ok,and i should have a lunch. 03:58:19 yuanyou_: ok 04:05:08 yuanyou_: this is the bundle i have please let me know if i am missing any option here http://paste.ubuntu.com/19997948/ 04:15:38 yuanyou_: u there? 04:44:36 narindergupta:hi,i am there 04:45:03 please check the bundle http://paste.ubuntu.com/19997948/ it looks good to me so i checked in the changes 04:45:39 and deployment is in progress 04:45:54 narindergupta: yes,so good to me 04:46:14 yuanyou_: good so looks like joid integratiion is completed then 04:47:01 narindergupta: yes 04:47:52 yuanyou_: cool. we have few congress issue due to that i can not run the scenario but when issue will get fixed i will run all scenarios again 04:48:38 narindergupta: ok,but can we create the scenario now. 04:48:49 yes i saw the patch 04:49:13 i think you need to add -d trusty in case of onos and sfc then we are good 04:49:41 you need to work with fdegir for the patch into erleng 05:01:37 narindergupta: ok, i will ask for fath to help add -d trusty 05:02:09 yuanyou_: that i already added but while sfc and onos you need to enable it 05:03:11 narindergupta: i don't know how can i enable it? 05:03:27 that you need to ask fdegir 05:03:39 how to pass the parameter during jenkins 05:05:16 narindergupta: which params we should pass? 05:05:36 -f sfc -d trusty 05:05:45 rest of them will be same 05:08:06 you mean this in jekens UBUNTU_DISTRO xenial 05:10:08 i have already defined the variable yes 05:10:57 but need to modify to trusty based on feature passed. 05:11:44 2.5.90-1 is the defalt openvswitch for all nodes currently and i am going to sleep now. It is 12:11 AM already here 05:12:06 narindergupta: ok,thanks 05:12:47 yuanyou_: good newa is deplyment completed so integration is good. 05:12:48 narindergupta: have a good night 05:13:30 Created a new network: 05:13:30 +---------------------------+--------------------------------------+ 05:13:30 | Field | Value | 05:13:30 +---------------------------+--------------------------------------+ 05:13:32 | admin_state_up | True | 05:13:32 | availability_zone_hints | | 05:13:35 | availability_zones | | 05:13:35 | created_at | 2016-07-19T05:12:59 | 05:13:37 | description | | 05:13:37 | id | 38af81d3-7a99-4377-a86d-322755c4b683 | 05:13:39 | ipv4_address_scope | | 05:13:39 | ipv6_address_scope | | 05:13:42 | is_default | False | 05:13:42 | mtu | 1450 | 05:13:44 | name | ext-net | 05:13:44 | provider:network_type | vxlan | 05:13:47 | provider:physical_network | | 05:13:47 | provider:segmentation_id | 1026 | 05:13:49 | router:external | True | 05:13:49 | shared | False | 05:13:51 | status | ACTIVE | 05:13:51 | subnets | | 05:13:54 | tags | | 05:13:54 | tenant_id | 0485cb4af2f94cf6ae6516f169080947 | 05:13:56 | updated_at | 2016-07-19T05:12:59 | 05:13:56 +---------------------------+--------------------------------------+ 06:16:11 fdegir: hi 11:55:37 David_Orange, hi David good morning how are you doing? 11:56:20 David_Orange, regarding this patch https://gerrit.opnfv.org/gerrit/#/c/17121/1 i have 2-3 comments if you have time we can go through together? 12:45:57 narinder: hi 12:46:10 David_Orange, hi how are you doing? 12:46:12 narinder: i am fine and you ? 12:47:27 narinder: i patch those 2 files this morning, hope it is ok for you, this is mainly cosmetic, but we can talk about that when you want 12:47:58 David_Orange, yes i saw that and i made a few comments 12:48:13 of course 12:48:20 for example deploy.py 12:48:35 yes 12:48:37 there was extra comma in comparison 12:48:47 at line 72 12:49:01 i think its not needed 12:49:37 ok, i check them all 12:49:40 in that file only the comment i have 12:49:53 in the genbunfdle.yaml 12:50:10 line 186 12:50:31 oh yes 12:50:35 sorry for elif :) 12:50:38 as liberty is supported in trusty only so i am checking trusty in extra if it is there then only support liberty 12:50:52 David_Orange, no need to sorry 12:51:34 my brain is not used to work with the hot here in Brittany ;) 12:52:15 is it very hot there? 12:52:58 in my opinion, the first part of deploy.py (the creation of opnfvcfg[]) should be set from an extra yaml file 12:53:59 very hot no :) probably really less hot than in texas, but we are not used to that :) 12:55:16 narinder: globally, are the changes ok for you ? 12:57:52 David_Orange, yes i am fine and not seeing any issue 12:58:35 David_Orange, did not understand this part the creation of opnfvcfg[]) should be set from an extra yaml file 13:00:21 narinder: i just think we should load 1 default yaml file or 2 to create opnfvlabcfg and opnfvcfg vars. 13:01:08 to split default vars and vars modification 13:02:06 currently i am reading from labconfig.yaml and converting deployconfig.yaml and deployment.yaml 13:02:18 and this happen during MAAS install time 13:02:20 but there are also a part that shall be loaded from labconfig.yaml (like vips) 13:02:30 ok 13:02:40 no vip i am creating based on network avialable 13:03:05 we should not take that input and it is standard like .21 to .30 13:03:39 i think more input asked from user always creates confusion and result in more dissatisfaction 13:04:07 ok, you removed the vip part i had set in labconfig.yaml 13:04:33 i understand, np. 13:04:44 no vip in labocnfig.yaml 13:04:56 we are generating it using deploy.py 13:05:15 and put into deployconfig.yaml during MAAS install time 13:06:05 i had set one, you remove it, this is not a problem. this is just that we have only ten vips with this solution 13:06:34 David_Orange, no problem we have 11-19 still empty and we can use that if needed 13:06:55 11-19 are used on orange pods 13:07:17 whats for? 13:07:26 for ipmi 13:07:39 ok no problem in that case 30-40 13:07:50 lets postone this part, we will see that when we will have no more ip 13:07:51 as i have started the dhcp from 50 13:07:59 np 13:08:15 David_Orange, sure we can decide as i kept lot of ips in range for 1-50 13:08:26 ok 13:09:41 i have just post a new version of the patch 13:12:56 David_Orange, line 186 13:13:05 still has issues 13:13:19 we need to ident it within if trusty 13:13:40 otherwise it will accept liberty with xeinal as well 13:13:50 is trusty is not provided which is invalid config 13:14:24 ok, i did not got that point 13:16:32 David those two ifs for liberty and trusty should be inclusive not exclusive 13:17:09 means trusty if should follow the liberty so if trusty is not passed then liberty option should not be checked 13:19:03 David_Orange, new patch was ok and I have merged it 13:21:07 narinder: I'm not seeing any reason that the line "connection = {{ shared_db.uri }}" in charm-congress/src/templates/liberty/congress.conf should be rendered as "connection = " in /etc/congress/congress.conf 13:21:52 narinder: you mentioned that there may be a layer change behind this. How would I find out what changes may have affected the rendering of {{ shared_db.uri }} ? 13:22:03 bryan_att, somehow shared_db.uri valus is not observed 13:22:27 bryan_att, we need to find out from openstack team 13:22:41 bryan_att, will you please consult with them? 13:22:43 what do you mean by "not observed" 13:23:13 bryan_att, means this value is empty so template converted as empty some how 13:23:49 where is the value supposed to be set? This is AFAICT a charm thing and not a Congress thing 13:25:25 i.e. in the design of the charm, what part of the charm takes the template and fills in the variables, and where does that get the values for the variables? This is specific to the charm and not congress. 13:31:14 bryan_att, i have no idea as I could not find the code who fills it and how. As layer concept is new and only few charms were written based on layers. 13:34:23 OK I will ask on #juju I guess 13:36:41 narinder: where do the "layers" get invoked? I suppose these are templates etc that are defined and maintained somewhere else and pulled in as needed by the build. IIUC, this causes a dependency which can break OPNFV deployments - so the dependency and how to address issues needs to be documented. 13:39:26 bryan_att, yes brian #juju sorry 14:09:39 David_Orange, looks like with public API congress code do not execute 14:10:00 David_Orange, so whenever you have time please have a look 14:10:21 ok, i will reinstall pod1 with latest version 14:10:37 i only have to specifiy publicapi as feature now, true ? 14:17:40 narinder: how do I redeploy congress on the POD to test out a charm patch? Can I make a local change or does it need to come from github? 14:17:56 bryan_att, yes you can do that 14:18:00 modify it 14:18:04 then update the charm 14:18:10 juju charm-update congress 14:18:18 should do 14:18:22 OK, where is the charm in the file system? 14:18:49 bryan_att, in intel pods it seems automatic deployment started 14:19:10 but it should be under ./joid/ci/xenial/charm-congress/ 14:19:13 has source 14:19:39 then you need to build it and move it to joid/ci/xenial/congress/ 14:19:53 then juju charm-update congress should work 14:20:28 there's no xenial folder in joid/ci 14:20:47 bryan_att, as i told installations restarted by CI 14:21:01 so everything wiped out and reployment started 14:21:11 OK, hopefully it will pick up the change 14:37:08 narinder: for publicapi install i just add the param in extrafeatures now, true ? 14:37:25 David_Orange, yes 14:37:39 David_Orange, everything else works except congress 14:37:54 ok, i will install pod1 and check 14:46:01 narinder: how do I know when the deploy is done so I can check it? 14:46:30 bryan_att, i will ping you 14:46:48 ok, is there some web page I can watch also? 14:47:24 bryan_att, current deployment may not have the fix but you can watch here https://build.opnfv.org/ci/job/joid-deploy-baremetal-daily-master/263/console 14:47:38 i am planning to start one right after that to pick the latest change 14:49:04 ok, I'll watch 15:08:52 narinder: i am lost with deployment method, joid-deploy.sh from releng is up to date ? what deploy script shall we use, deploy.sh or .py ? 15:09:09 i use deploy.sh 15:09:23 that use deploy.py ? 15:09:57 maas uses deploy.py 15:10:05 ok 15:10:12 and create the requiredfile which i copy in ~/.juju/ 15:10:30 maybe a file rename can be a good thing to avoid confusion 15:11:01 then those will be used in deploy.sh to start the deployment and 01-deploybundle.sh really calls genbundle.py 15:11:12 i will update joid-deploy.sh to enable extrafeature and remove old things (password....) 15:11:19 ok 15:14:15 sure that will help if uses the joid-deploy.sh 15:14:38 also i use deploy.sh and now you can pass mutiple feature "," seperated 15:14:52 like -f sfc,ipv6,trusty 15:15:56 ok so you mix nfvfeature and extrastuff ? 15:16:07 or extrastruff is another param ? 15:19:04 David_Orange, from deploy.sh all feature can be passed using -f 15:19:15 and i do not have seperate parmameter for that 15:19:41 but while passing argument to genBundle i take care of it based on parameter passed 15:19:52 hope it make sense for you 15:22:35 no offense, but it is confusing for me 15:25:23 give me 5 minutes 15:25:53 David_Orange, ok 15:29:10 sorry got a phone call 15:30:56 for me it seems complicated: we give scenario name to jenkins > we split it to give all features (mixing it with extra) to deploy.sh > we recreate a scenario name to give it to genbundle 15:32:09 i think giving scenario name to all sript would be easier to understand 15:34:18 for publicapi, you does not get it from scenario name, but you get it from os-domain-name in deployconfig.yaml 15:41:31 narinder: for features param, we split it with '_' from scenario name and create a new one for deploy.sh with ',' ; from my point of view it is confusing. But again, no offense, i do not have all the scripts history and the job is done 15:42:33 but if we decide to use deployconfig.yaml for passing params, this is not a problem 15:43:09 David_Orange, deploy.sh we already implemented feature using "," and few scenario name has odl_l2 and odl_l3 which is breaking the current deployment 15:43:21 so i think it should not be a problem 15:45:48 i think, to have something more homogeneous, is to get nfvfeatures and extra in joid-deploy.sh, pass all to deploy.sh, generate deployconfig.yaml, and read it in genbundle 15:46:42 it will minimize changes in deploy script and will only impact joid-deploy.sh and genbundle.py 15:47:34 narinder: it will just remove scenario name creation before calling genBundle 15:47:56 so from a joid perspective, scenario name is never used 15:51:02 David_Orange, once you submit a patch i will have a look and feature passed to deploy.sh gets converted into scenario to create the bundles using genbundle.py 15:51:15 so change has to be done very carefully 15:51:41 and reason i did was to have compaitibility in deploy.sh and genbundle.py 15:54:39 i fully understand 15:55:53 narinder: i will check if i can change genbundle and check if i can get all params from deployconfig.yaml (it is generated before genbundle exec, true ?) 15:56:26 narinder: looks like the job finished but congress did not get installed fully. how do we start it again to make sure it gets the latest charm? 16:00:17 bryan_att, try this https://build.opnfv.org/ci/job/joid-deploy-baremetal-daily-master/266/console this is just started after redeploy 16:01:06 deloyconfig.yaml craeted during MAAS install before the genbundle exec yes 16:01:18 and that will be based on labconfig providede 16:01:23 for each lab 16:02:10 so if i can find all informations included in scenario name, we do not need scenario name in genbundle 16:04:50 David_Orange, unfrotunatel it is not possible because after maas deployment you can chose any scenarios 16:05:02 which can include mutiple or single feature 16:05:27 i am fins scenario name in genbundle as i modified by script accordingly 16:05:42 you can see in 01-deploybunle.sh i am changin , to _ 16:06:06 also appending the scenario accordingly 16:06:34 true, but for publicapi extra feature, you do it only if the param is present in the deployconfig.yaml, so it will be always, not linked with scenario name 16:06:37 i know we can combine feature and extra into one but since we have already codede accoridngly so no need to worry for now 16:06:46 correct 16:06:59 which is true right and thats how it should work 16:07:42 why, that way it is not linked with extraparam, but labconfig 16:08:19 David_Orange, if lab is defining the publicapi then bundle should use 16:08:26 otherwise not 16:08:58 extraparam will be passed to genbundle based on labconfig definitions 16:09:15 so that everything is automated 16:09:33 if you can look into the ornage pod1 labconfig.yaml i have change it accoridngly 16:09:45 yes i see 16:12:04 narinder: i understand how it works, that way if we remove publicapi in jenkins scenario name, it has no impact on deployment, except if we redeploy maas, true ? 16:14:05 you do not have to reploy maas 16:14:14 just change deployconfig.yaml in ~/.juju 16:14:33 and the publicpai wont be use after the MAAS deployment 16:14:44 yes, this is the manual way, but from jenkins, it wont works 16:14:49 but not through jenkins though. 16:15:01 it can 16:15:11 yes you are true it won't 16:15:24 i can remove this from my script if you want 16:15:41 so that you can control from jenkins 16:16:03 but my main worry is if this value is not deifned then we have some random values 16:16:22 please dont, you probably have more important to do 16:16:42 which value ? 16:17:15 publicapi value 16:17:23 publicapi is just an example of the problem of rebuilding the scenario name from labconfig 16:18:22 not labconfig bu deployconfig 16:18:30 labconfig i am usign to get the values for automation 16:18:56 we need parameters and asking from user is not a good idea 16:19:16 its better put the smartness in deployer rather than in jenkins 16:20:01 thats the input i got when we review this with MArk S earlier 16:21:32 we can add this as feature and pass it using -f but i chose not to because it is lab specific value 16:21:32 i fully undestand that point and totally agree, but have something that seems enabled from jenkins that will not will be a problem 16:22:27 i agree from jenkins it should be modifiable. LEt me see my code and see how can we pass through jenkins 16:22:42 as well 16:22:55 the other solution is not to use extra feature part of the scenario name and modify it in labconfig 16:24:20 i am using osdomainname: pod1.opnfv.fr in labconfig 16:24:30 if this is not deinfed then i am not using it 16:24:34 otherwise yes 16:25:02 so question is if we do not want to use this even though it exist 16:25:05 correct 16:25:33 it exist but is optional 16:27:06 in genbundle i used extra to enable beta features, to fully test a beta feature with all the CI chain 16:27:52 i though it can be a good thing to do before pushing this feature from extra to nfvfeature 16:31:31 David_Orange, currently i am passing as extra feature only to genbundles 16:31:47 var=os-$4-$fea-$1"-"$6"_"$2 16:31:56 where $4 is sdn 16:32:13 $fea is feature requested 16:32:37 and $1 is type 16:32:54 and $6 is distro 16:33:44 and $2 is labname 16:34:05 and after that based on osdomname i am adding _publicapi which is extra feature 16:34:10 so you reuse the jenkins variable and change it 16:34:40 my joid is not depend upon jenkins 16:34:47 joid code is not 16:35:17 for publicapi i uses domname if exist then pass extra feature to genBundle otherwise not 16:35:29 yes i know, but from a joid user/developper it can be confusing 16:35:41 why? 16:36:13 joid user should not know about this right. They should know domname of their lab 16:36:24 and put into labconfig.yaml thats it 16:36:31 rest joid code will do 16:38:03 joid user does not haer to know that, but joid developpers must. to have more contribution, the way we pass args from scripts to scripts is important 16:38:35 David_Orange, yes i agree 16:38:47 if you are ok, i will rename "scenario" in genBundle to have something different 16:39:00 i think i shoudl write more documention or comments on it 16:39:16 sure you can 16:40:06 yes :) and document how to enable features on deployconfig: "To enable public API, please fillup osdomname" 16:41:51 i though i need more spare time to do that. sorry for all confusion 16:43:13 narinder: i will make a workflow diagram if you agree, it will give me more understanding on how the deployment is done, and i will share it to you and all devs 16:44:16 please do that if you can 16:45:26 ok, i will try 16:45:56 thanks 16:46:20 but for your info 02-maasdeploy.sh will deploy maas and should execute first. 16:46:30 then 00-bootstrap.sh 16:46:43 then 01-deploybundle.sh 16:46:44 i have to go, see you tomorrow (i will probably can not attend to joid meeting, i keep you in touch) 16:46:53 see you 16:47:53 02>00>01 is also confusing, but we will se that later :) 18:36:11 bryan_att, looks ilke that fixes the congress issue. I believe you should look now moving this to openstack-charm 18:36:55 bryan_att, jamespage mentioned this exmaple to start the work https://review.openstack.org/#/c/339485/ 18:37:13 i think you need to modify it and submit a new pull request 19:12:32 narinder: for some reason the congress charm did not work on my NUC lab. did you see it work somewhere else? 22:38:27 narinder: I see now you had a patch to comment out the congress relations. You removed this after I cloned this AM. But I have another issue. Why is the region "Canonical" now when in the past (and in many things under the JOID repo) it was "RegionOne". I can't access the congress API now because of the region difference. 22:38:38 narindergupta: ping 22:39:25 bryan_att, pong i always had Canonical. 22:39:41 bryan_att, i never used ReionOne 22:39:46 why the difference now? 22:40:08 bryan_att, its passing in my deployment 22:40:33 bryan_att, https://build.opnfv.org/ci/job/joid-deploy-baremetal-daily-master/268/console 22:48:53 narindergupta: any idea why it would be different for me? https://www.irccloud.com/pastebin/UwkzmCxE/ 02:21:51 narindergupta: looks like the RegionOne thing was an side-effect of running the "remove-service congress" and then running "juju-deployer" to redeploy it. I re-cloned and installed from scratch and it's as you said - all endpoints are in the Canonical region. 02:39:01 May-meimei: ping 02:52:22 yuanyou_, hi 02:55:38 bryan_att, cool and thanks for confirmation 03:05:47 yuanyou_: ? 03:17:39 narindergupta: hi 03:18:12 May-meimei: hi,can you submit releng https://gerrit.opnfv.org/gerrit/#/c/16875/ 03:18:22 yuanyou_, so whats the news were you able to enable the scenario? 03:19:28 narindergupta: I had send an email to fdegir to help add enable the scenario 03:20:06 yuanyou_, ok 03:20:18 narindergupta: i am waiting for him 03:20:47 yuanyou_, ok let me know once ready. 03:21:23 narindergupta: may be May-meimei can do this also, I will notice you first. 03:21:45 submited 03:21:50 ok 03:22:04 May-meimei: cool 03:22:20 senario will be availble later 03:22:56 May-meimei: but i need params -f sfc -d trusty to pass 03:23:46 deploy.sh -f sfc -d trusty ? 03:23:51 May-meimei: yes 03:25:32 but default value is mitaka, yes? 03:28:37 why not add a branch in joid-deploy.sh ? 03:29:43 May-meimei: mitaka is -o 03:30:54 oh, yes, confused 03:33:44 May-meimei: when the scenario is joid-os-onos-sfc-ha, run ./deploy.sh -s onos -d trusty -f sfc -t ha -o mitaka 13:55:28 narinder: hi 14:01:21 narindergupta: how are you ? 14:01:41 David_Orange, hi i am fine 14:01:45 what about you? 14:01:56 narindergupta: i am fine, thx 14:02:05 i have a juju question for congress 14:02:52 is there a way to force juju tu use the default value and not the value given in the top bloc 14:03:56 os-public-hostname is set for all the bundle, and to make congress works internaly we need to set this value to its default, not the one set at the top, do you follow me ? 14:04:37 if i can not, i have another way, but it will be a little more longer for the post install script 14:04:46 David_Orange, yes only way to do is define it for individual charm all option needed 14:04:57 ok 14:05:21 need to remove the top values and insert place it for individual charm 14:05:39 so it is as you want: 14:05:39 - we set the value for all services, except congress 14:05:39 - wet change the value in post install 14:05:46 with the current way we bundle it it is quite possible 14:05:57 correct 14:06:22 David_Orange lets do it for individual service 14:06:34 soecially for public api 14:06:43 as we have a way to do it now 14:07:22 ok, so i patch all services with a 'if publicapi: os-public-hostname=xxxx' except for congress 14:07:46 correct 14:07:53 and also we do not have to patch all 14:08:00 only whereever applicable 14:08:11 for example promise may not need it 14:08:16 ok , i do it, i hope i wont forget a service 14:08:24 nova-compute may not need it 14:08:40 sure all service the have that parameter 14:11:57 not all 14:12:03 nova-compute does not have it 14:12:18 sorry it does 14:12:27 dont worry i will check 14:12:36 ok 14:12:41 no it does not actually 14:12:47 confress has it 14:12:53 another point: in ceph.yaml template 14:12:59 yeah 14:13:11 line28 14:13:43 let me open that file 14:14:09 ok 14:14:22 its checking the mode ha 14:14:38 yes, i did not understand, but now it is ok 14:14:50 for ceph we need a minimum of 2 14:15:01 but more if in ha mode 14:15:04 minimum is 3 14:15:27 but for non ha since we are doing deployment with 2 nodes only so we have only 2 14:17:09 the idea is to deploy on all available nodes, is it correct ? 14:18:29 yes 14:18:40 but some lab has 2 nodes only 14:18:47 is there a place where the pod size is set, or is it harcoded ? 14:19:05 no i am setting ti based on nodes defined in labconfig.yaml 14:19:25 i am counting it and set in deployconfig.yaml 14:19:48 opnfv: units: 5 14:19:58 in case of intel pods 14:20:19 ok, this could be a good thing to get that value here for ceph 14:20:48 definetely some logic is needed for nonha as we need to have more nova-compute as well in that case 14:21:11 so that we can utilize all resources 14:21:32 whether HA or nonha 14:23:45 i will see later how to get it in genBundle 14:24:23 1/ template modify for congress 14:24:23 2/ workflow diagramm 14:24:23 3/ genBundle modify depending on workflow analysis 14:24:48 agreed lets create a jira also for same 14:25:00 so that we wont miss any work item 14:25:07 ok for me 14:25:12 ok will do it now 14:25:50 i let you do that, i have only 30 minutes available for congress, and i have a meeting on another subject 14:26:13 i can not connect to joid meeting, sorry 14:27:41 no problem 14:27:58 yeah send me patch and i will restart the deployment on orange pod1 14:29:21 please wait before changing pod1, valentin is working on it for public api 14:29:43 oh ok 14:45:23 is there a jira for congress ? 14:49:12 there was to integrete 14:49:23 ok 14:49:25 https://gerrit.opnfv.org/gerrit/17235 14:50:50 do we need use-internal-endpoints: true also for all or that will work? 14:51:02 as being override 14:53:12 ok i am +2 it and merging it if still find out issue we will fix it further 15:00:48 narinder, it will work with interal everytime 15:00:54 ok 15:01:08 i have to go, see you tomorrow 15:01:40 sure 15:01:52 i have merge the changes 16:04:17 #info Dave Urschatz 16:05:01 arturt: Error: Can't start another meeting, one is in progress. Use #endmeeting first. 16:05:19 #endmeeting