Friday, 2016-08-26

*** zhurong has quit IRC00:04
*** elynn has quit IRC00:35
*** zhurong has joined #senlin01:17
*** yanyanhu has joined #senlin01:27
*** elynn has joined #senlin01:52
*** elynn_ has joined #senlin01:57
*** elynn has quit IRC02:00
*** guoshan has joined #senlin02:11
*** guoshan has quit IRC02:26
openstackgerritMerged openstack/senlin: api-ref: OpenStack-Request-Id and location
yanyanhuhi, Qiming, around?02:38
yanyanhuabout the subscriber url, I'm considering to add a new receiver API interface, e.g. /v1/receivers/{receiver_id}/notify02:39
yanyanhuand also a new engine service call receiver_notify02:39
yanyanhualso add a simple description about trust usage in subscription to this etherpad:
yanyanhujust read the code of zaqar's trust notification driver, I think we can apply this way to support authentication for receiver notify API access02:42
Qimingso the receiver-notify will only server authenticated and authorized requests?02:44
yanyanhuand it will be invisible for enduer02:46
yanyanhuwill make some tries on this design to see whether it can work as expected, especially the trust subscription part02:49
openstackgerritQiming Teng proposed openstack/senlin: Fix docs about autoscaling
*** yanyanhu has quit IRC02:56
*** yanyanhu has joined #senlin02:58
openstackgerritNam Nguyen Hoai proposed openstack/senlin: Clean imports in code
openstackgerritMerged openstack/senlin: Improve server metadata format
*** gongysh_ has joined #senlin03:59
openstackgerritMerged openstack/senlin: Fix docs about autoscaling
openstackgerritQiming Teng proposed openstack/senlin: Fix api version mismatch return code
*** xuhaiwei has quit IRC04:11
*** elynn_ has quit IRC04:31
openstackgerritXueFeng Liu proposed openstack/senlin: Fix policy command example
*** elynn_ has joined #senlin04:56
*** elynn_ has quit IRC05:00
*** elynn_ has joined #senlin05:01
*** elynn_ has quit IRC05:06
*** elynn_ has joined #senlin05:08
openstackgerritOpenStack Proposal Bot proposed openstack/senlin: Updated from global requirements
Qiming$ senlin cluster-run -u cirros -n private -t fixed -s c105:34
Qimingnode: f9274ccb-99fa-409d-8678-37dfd2645b8505:34
Qimingstatus: SUCCEEDED05:34
Qimingnode: cdd5c318-2c85-4cf8-be2b-fd353a2de8be05:35
Qimingstatus: SUCCEEDED05:35
Qiming$ cat r1.sh05:36
Qimingls /05:36
Qiming$ senlin cluster-node-list c105:36
QimingWARNING (shell) "senlin cluster-node-list" is deprecated, please use "openstack cluster node members list" instead.05:36
Qiming| id       | name              | index | status | physical_id | created_at           |05:36
Qiming| cdd5c318 | node-db59d9de-004 | 4     | ACTIVE | 342b31c1    | 2016-08-26T05:21:16Z |05:36
Qiming| f9274ccb | node-db59d9de-003 | 3     | ACTIVE | 4fb20876    | 2016-08-26T05:21:18Z |05:36
openstackgerritMerged openstack/senlin: Close health manager loop
openstackgerritOpenStack Proposal Bot proposed openstack/senlin: Updated from global requirements
openstackgerritYanyan Hu proposed openstack/senlin: Rework receiver delete
yanyanhuQiming,  cool :)06:05
openstackgerritYanyan Hu proposed openstack/senlin: Rework receiver delete
lixinhuiQiming, Looks good06:47
lixinhuihave you read the code of Nova about host failure part06:47
Qimingit was a client-only patch06:47
lixinhuias far as I know06:47
Qiminghaven't figured out how to transfer private key securely to senlin server and let the server do the ssh later06:48
lixinhuiNova only provide heartbeat nowdays06:48
lixinhuiyes, Qiming, that is the critical part06:48
lixinhuiwhat I am asking is about next step of host failure fencing06:48
Qimingokay, have to dig the openstack-dev mailinglist06:49
Qimingapril 0706:51
Qimingthis was the response:
lixinhuiI think I can understand this thread from my reading of nova's code about service status notifinication part06:57
lixinhuiWhen admin trigger host into Maintenance, he/she is supposted to disable the nova services firstly06:58
lixinhuiand accordingly, the event will be sent out06:58
Qimingaccidental compute down is actually detected, but not reported06:59
Qimingthat was my impression06:59
lixinhuithat does not mean any event will happen when heartbeat failed, heartbeat still only change the DB06:59
lixinhuiyou can confirm07:00
lixinhuithen discuss what we should do for host failure fencing07:01
Qimingideally, nova knows the status of nova-compute07:09
Qimingwhen a nova-compute service is down, that physical machine is useless07:09
Qimingit should emit a huge notification (or in separate steps) containing the host info, hypervisor info, and the info about the list of vm instances running on that machine07:10
Qimingyanyan, any idea why 360911 fails?07:11
QimingI don't think nova has that capability today07:14
yanyanhuQiming, oh, didn't notice it failed, let me check it07:18
Qimingyou forgot to add test case to the change of the service module07:18
yanyanhuah, yes07:19
yanyanhuwill add it07:19
yanyanhuBTW, about get_endpoint support in sdk session
yanyanhuI made a test and found region filtering doesn't take effect07:19
yanyanhuwhile service_type and interface filtering works07:20
openstackgerritMerged openstack/senlin: Fix api version mismatch return code
yanyanhuQiming, so may need a fix at sdk side?07:33
Qimingline 225 is invoking keystoneauth107:33
Qimingyou mean the filter is not working there?07:33
yanyanhuyes, seems so07:33
yanyanhuI passed region=$REGION_NAME as kwargs07:34
yanyanhubut I found kwargs was not used in get_endpoint07:34
Qimingin sdk, you set region via profile, right?07:34
Qiminglook at line 22207:35
yanyanhuyes, but the region of service I want to get endpoint for could be different from the one defined in profile I think07:35
Qimingyou are supposed to do profile.set_region(...) before that?07:35
Qimingit may and may not make sense always, right?07:36
yanyanhuQiming, sometimes, maybe, e.g. services from different regions registered to one keystone07:36
Qiminghow would you assume keystone at one region knows the endpoint for all other regions?07:36
Qimingyes, that is possible07:37
yanyanhuso in that case,  I think this won't work?07:38
Qimingto support that case ... you may have to fix sdk07:38
yanyanhuumm, guess so07:38
yanyanhunot sure whether keystoneauth1 provides this support07:39
openstackgerritYanyan Hu proposed openstack/senlin: Rework receiver delete
Qimingabout keystoneauth1 support, you may want to try a quick fix07:42
Qimingpass the kwargs directly to keystoneauth.session.get_endpoint()07:42
yanyanhuQiming, yes07:42
yanyanhulet me have a try07:42
Qimingthe base authentication supports this:07:43
Qiming159     def get_endpoint(self, session, service_type=None, interface=None,07:43
Qiming160                      region_name=None, service_name=None, version=None,07:43
Qiming161                      allow={}, **kwargs):07:43
yanyanhuoh, I guess that means region_name based filtering is supported07:43
yanyanhuhi, Qiming, if I add 'region' key to filt, it works07:57
yanyanhue.g. after line 224, add if region: filt.region = region07:57
yanyanhubut just as you said, not sure this is real use case07:57
yanyanhufrom sdk's perspective07:58
Qimingyou may want to add a region_name=None to the parameter list08:00
Qimingthen the key on line 218 becomes (service_type, region_name, interface)08:00
yanyanhuif so, for the case using cache, region_name will be used to filter endpoint08:03
yanyanhuhi, Qiming, do we need to propose this kind of fix to sdk?08:04
Qimingi think it makes sense08:06
Qiminglet's see if brian agrees to it08:06
yanyanhuI see.08:06
yanyanhuwill propose a patch for it08:06
yanyanhuanyway, it won't be included into newton release. But I think it's ok us08:07
yanyanhuok for us08:07
Qimingwell, that is a good point08:07
Qimingmaybe we should prepare for both08:07
Qimingideally, the logic should go to sdk08:08
yanyanhuI see08:08
Qimingbut if we are trying to retrieve service endpoint, that version based endpoint select logic is the right thing to do08:08
Qimingif we want this feature to be included in senlin newton release, we can keep a local copy of the logic08:09
Qimingand then remove it when sdk has the fix08:09
yanyanhuyou mean overriding the get_endpoint of sdk session?08:09
yanyanhuoh, we just implement the version match logic in the get_base_url method?08:10
yanyanhulike what session.get_endpoint does08:11
yanyanhurather than using hard coded 'v1' there08:11
*** openstackgerrit has quit IRC08:18
*** openstackgerrit has joined #senlin08:19
*** ChanServ sets mode: +v openstackgerrit08:19
Qimingavoid hardcoding would be great08:21
Qimingalthough we don't know for sure when we will have to upgrade to v208:21
openstackgerritQiming Teng proposed openstack/python-senlinclient: Add 'cluster-run' command
openstackgerritOpenStack Proposal Bot proposed openstack/senlin: Updated from global requirements
*** elynn_ has quit IRC09:29
openstackgerritYanyan Hu proposed openstack/senlin: Minor tweak webhook receiver
openstackgerritMerged openstack/senlin: Enable health policy to handle CLUSTER_SCALE_IN
openstackgerritYanyan Hu proposed openstack/senlin: Add version support to schema and spec
openstackgerritlvdongbing proposed openstack/senlin: Fix reraise issue
*** zhurong has quit IRC10:01
*** yanyanhu has quit IRC10:02
*** zzxwill has joined #senlin10:03
*** yanyanhu has joined #senlin10:03
*** yanyanhu has quit IRC10:03
openstackgerritMerged openstack/senlin: Updated from global requirements
openstackgerritxu-haiwei proposed openstack/senlin: Fix some bugs for container creation
*** elynn_ has joined #senlin11:45
*** elynn_ has quit IRC11:48
*** zzxwill has quit IRC11:49
*** elynn_ has joined #senlin11:51
openstackgerritQiming Teng proposed openstack/senlin: Update api-ref configuration
*** zhurong has joined #senlin13:04
*** zhurong has quit IRC15:03
*** Drago has joined #senlin15:47
*** Drago has quit IRC15:47
*** Drago has joined #senlin15:47
*** elynn_ has quit IRC16:15
openstackgerritxu-haiwei proposed openstack/senlin: Fix some bugs for container creation
*** Drago has quit IRC17:26

Generated by 2.14.0 by Marius Gedminas - find it at!