Monday, 2017-03-20

*** catintheroof has joined #senlin00:20
*** catintheroof has quit IRC00:26
*** elynn has joined #senlin01:18
*** elynn_ has joined #senlin01:23
*** elynn has quit IRC01:25
*** guoshan has joined #senlin02:12
*** XueFeng has joined #senlin03:10
yuanbinelynn_, about senlin security_groups format in cirros_basic.yaml, I change how to write this file ?03:23
yuanbinelynn_, I use this file
*** zhurong has joined #senlin03:25
yuanbinelynn_, But the security_groups will be list, if security_groups:3acfdc18-22e0-4ff6-a153-4db846f5da18, the code will be change to ['3','a','c','f',...] format03:26
openstackgerritRUIJIE YUAN proposed openstack/senlin master: failed to dump 'WARNING' notification
elynn_Hi yuanbin03:38
elynn_security_groups is a list03:38
elynn_Use security_groups: ["xxx"]03:38
openstackgerritRUIJIE YUAN proposed openstack/senlin master: failed to dump 'WARNING' notification
elynn_For profile type os.nova.server , networks didn't support security_groups yet.03:41
elynn_yuanbin, ^03:41
Qimingelynn_, I think he was asking a different question03:55
Qiminghis question is mainly about how to describe a list in YAML03:55
elynn_Qiming, I see... I was thinking is it okay to introduce a strlist schema, which accept a string or a list, but return a list.03:57
elynn_I was thinking to use it at 'fixed_ip', and let it accept a list or a string without breaking compatibility .03:59
*** guoshan has quit IRC04:00
*** XueFeng has quit IRC04:07
openstackgerritMerged openstack/senlin master: Added CORS support to Senlin
*** zhurong has quit IRC04:16
*** elynn_ has quit IRC04:17
openstackgerritliyi proposed openstack/senlin master: Add access log for devstack.
*** zhurong has joined #senlin04:59
*** elynn has joined #senlin05:03
*** elynn_ has joined #senlin05:07
*** elynn has quit IRC05:08
*** zhurong has quit IRC05:14
openstackgerritRUIJIE YUAN proposed openstack/senlin master: failed to dump 'WARNING' notification
*** ruijie__ has quit IRC05:27
*** XueFeng has joined #senlin05:38
*** ruijie has joined #senlin05:53
ruijiehi, got a senlin client problem, not sure if this is a bug05:56
ruijieplz see the 2nd one... sorry :(05:59
elynn_Create cluster failed?06:10
ruijieyes elynn_06:11
ruijielooks like client failed to parse the err msg06:11
elynn_FINAL-TEST is cluster name?06:12
ruijieits the profile name06:12
elynn_Seems like sdk failed to parse the error.06:13
*** ruijie_ has joined #senlin06:13
Qiminglooks like a SDK version06:13
elynn_messages = [obj.get('message') for obj in resp.json().values()]06:14
elynn_Obj here seems like a unicode06:14
Qimingcheck line 12906:14
Qimingit was added recently06:14
Qimingthis logic was broken by a developer when adding some neutron specific error parsing into sdk06:14
*** ruijie has quit IRC06:16
elynn_It's in 0.9.14, so I guess latest version won't break it.06:16
elynn_BTW, I failed to enable senlin-dashboard following the guide of senlin-dashboard.
openstackLaunchpad bug 1674216 in senlin "senlin dashboard can not enable" [Undecided,New]06:17
ruijie_I am using 0.9.13, let me update it and check it again06:17
XueFengdashboard has bug recently06:23
ruijie_0.9.14 is fine :)06:23
elynn_XueFeng, is this bug that lead to the failure I met?06:24
*** guoshan_ has joined #senlin06:25
*** guoshan has joined #senlin06:29
*** guoshan_ has quit IRC06:29
XueFengWe discuss two problem. One is ruijie report, this is cause by sdk.06:37
XueFengAnd it maybe only the problem when exception handling.06:38
XueFengAbout dashboard, which I met is a install problem.06:40
XueFengWhen I enable senlin-dashboard, I can't install devstak sucessfully06:40
*** 7IZAATF07 has joined #senlin06:42
*** 7GHAALN3P has joined #senlin06:42
*** 7GHAALN3P has quit IRC06:42
*** guoshan has quit IRC06:45
*** guoshan has joined #senlin06:45
elynn_XueFeng, How do you enable senlin-dashboard?07:13
elynn_enable_plugin senlin-dashboard https://xxx/senlin-dashboard ?07:13
openstackgerritliyi proposed openstack/senlin master: Add access log for devstack.
openstackgerritliyi proposed openstack/senlin master: Add access log for devstack.
elynn_Hi Qiming , around?07:32
*** Jeffrey4l__ has quit IRC07:33
*** Jeffrey4l__ has joined #senlin07:36
*** elynn__ has joined #senlin08:02
elynn__Qiming,  I'm about to back port vdu , should I create a new server profile and set type as os.nova.server-2.0?08:03
*** elynn_ has quit IRC08:04
elynn__Or just override the current server profile?08:04
Qimingem, I'd suggest we just add properties to the existing one08:29
Qimingbut you will bump the profile version in place08:30
elynn__Okay, but if user use os.nova.server-1.0, will we break it?08:30
Qimingyou may want to check this:
Qimingfor schemas, we have min_version and max_version support08:31
Qimingalthough not complicated as versioned objects, we do have versioning support08:32
*** XueFeng has quit IRC08:32
QimingA spec has a version, you will compare the schema version against it to determine if a property is acceptable08:33
Qiminge.g. nova server 1.0 will not accept a new property if you have labeled property with min_version 1.108:33
elynn__Okay, I see how it work now...08:35
ruijie_Hi Qiming around?09:01
*** openstackgerrit has quit IRC09:02
ruijie_is this process too complex for us ?09:04
ruijie_as far as I concerned, line 432-466 is for creation process09:05
ruijie_and the rest is for scaling, resizing and recover09:05
QimingI know this project, but I have never read its code09:06
QimingI'm not quite sure how otter manages that tasks it is building09:06
Qimingsimilar to the "action" concept in senlin?09:06
ruijie_sorry ... I saw it here ..
ruijie_I am trying to let our resizing and recovering process support desired_capacity and health check09:07
Qimingem ...09:09
QimingI'm a little bit conservative about convergence09:09
Qimingif I'm understanding it correctly (at least in Heat), convergence is a word spaning three different jobs09:10
ruijie_em ... if we do these processes during scaling and recovering ... I am actually worrying about the performence..09:10
Qiming1. make things parallel, this may be irrelevant in our current dicussion context09:10
*** openstackgerrit has joined #senlin09:10
*** ChanServ sets mode: +v openstackgerrit09:10
openstackgerritMerged openstack/senlin master: Add access log for devstack.
Qiming2. make sure the engine is can be made aware of the actual status of physical resources WHEN REQUIRED09:11
openstackgerritMerged openstack/senlin master: failed to dump 'WARNING' notification
openstackgerritMerged openstack/senlin master: Add get_stack_template support for heat
openstackgerritMerged openstack/senlin master: Add get_env and get_files function to heat driver
Qimingthis should not be a problem for nodes and clusters today, although I'm not 100% percent confident about its stability09:11
Qiming3. derive a plan to converge the current states of resources to the "desired" states ... this is my top concern09:12
Qimingactually, the health policy is trying to introduce such a loop in to Senlin engine09:12
Qimingdetection -> (notification ->) recovery09:13
ruijie_yes Qiming, they do status check first if needed, and then based on the status of nodes to get the nodes needed to be created, updated or deleted09:13
Qimingmy concern is mainly related to the reliability of detection and recovery09:13
Qimingif an environment is not reliable, we will get false alarms, mistakenly treating some healthy resources as unhealthy09:14
ruijie_em, can we just do size check in resizing process and let recover action do status check09:14
Qimingthis can be caused by network delay, missing notification, busy service ... for example09:14
Qimingem, you can never get a perfect solution in a very dynamice environment09:15
Qimingsay you have just completed a status check, just before you started to recover some bad nodes, another 3 nodes are going offline09:16
Qimingshould we handle them at all?09:16
ruijie_yes Qiming, so what we can do is finding relatively good ones09:16
QimingI'd prefer making them configurable09:16
Qimingor ideally, orthogonal09:16
ruijie_Qiming you mean for both resizing and recovery ?09:17
Qiminglet health be health, let scale be scale09:17
Qimingif you don't want to apply health policy, and you still want to do a status check before doing scaling, fine09:17
QimingI have no objections on that09:17
Qimingbut I'd warn you that detection node failure is very tricky09:18
QimingICMP is not trustworthy, heartbeat can be lost, Nova API may not be so responsive, ...09:18
Qiminga more reliable way to detect node failure is to inquire the Load Balancer09:19
Qimingbut, as you may know, Ocativia is constantly rejecting our patches ...09:20
Qimingeven they have some good points, we failed to follow up  ...09:20
ruijie_sign ..09:20
Qimingso ... in this community09:20
Qimingjust as someone mentioned  on mailing list09:20
Qimingit is do-mocracy, rather than democracy09:21
Qimingneed some features, some business requirements, very good usage scenarios?09:21
Qiminggood, jump on to it and get it implemented09:21
Qimingno one will do that for you , :)09:22
Qimingwe really hope ocavia can be stable and usable soon09:23
ruijie_that is ..09:23
Qimingbut ... there is no way to drive it,  unless we jump onto it ourselves09:23
ruijie_anyway, we can focus on our project, if possible, maybe we can implement it ourselves ..09:24
Qimingsee this:
ruijie_that is also common even in one company :(09:25
Qimingno one answers your question, :)09:25
ruijie_that is unreasonable ... :)09:28
ruijie_And for this support of health check and desired capacity check, I want to file a spec for it09:29
ruijie_So that we can also discussed it offline09:30
*** guoshan has quit IRC10:27
*** 7IZAATF07 has quit IRC11:07
*** yanyanhu_ has joined #senlin11:09
*** yanyanhu_ has quit IRC11:09
*** zhurong has joined #senlin11:17
*** elynn__ has quit IRC11:24
*** elynn__ has joined #senlin11:28
*** elynn_ has joined #senlin11:32
*** elynn__ has quit IRC11:32
*** Jeffrey4l_ has joined #senlin11:34
*** Jeffrey4l__ has quit IRC11:36
*** zhurong has quit IRC11:37
*** elynn_ has quit IRC12:39
*** catintheroof has joined #senlin12:57
openstackgerritJeffreyGuan proposed openstack/senlin stable/mitaka: update the typo in setup.cfg from "senlin.common.wsgi = senlin.common.wsgi:list_opts" to "senlin.api.common.wsgi = senlin.api.common.wsgi:list_opts"
*** guoshan has joined #senlin13:39
*** guoshan has quit IRC13:58
*** guoshan has joined #senlin14:06
openstackgerritJeffreyGuan proposed openstack/python-senlinclient stable/mitaka: Fix the type error for "result" in senlinclient.
openstackgerritJeffreyGuan proposed openstack/senlin stable/mitaka: update the typo in setup.cfg from "senlin.common.wsgi = senlin.common.wsgi:list_opts" to "senlin.api.common.wsgi = senlin.api.common.wsgi:list_opts"
*** guoshan has quit IRC15:24
*** catintheroof has quit IRC18:08
*** catintheroof has joined #senlin18:08
*** zhenguo_ has joined #senlin20:18
*** chohoor_ has joined #senlin20:24
*** chohoor has quit IRC20:29
*** zhenguo has quit IRC20:29
*** zhenguo_ is now known as zhenguo20:30
*** Jeffrey4l_ has quit IRC21:35
*** Jeffrey4l_ has joined #senlin21:38
*** chohoor_ has quit IRC21:39
*** chohoor has joined #senlin21:40
*** catintheroof has quit IRC23:13

Generated by 2.14.0 by Marius Gedminas - find it at!