| *** shu-mutou-AFK is now known as shu-mutou | 00:00 | |
| *** Qiming has quit IRC | 00:04 | |
| *** sridhar_ram has quit IRC | 00:38 | |
| *** Qiming has joined #senlin | 01:09 | |
| *** sridhar_ram has joined #senlin | 01:35 | |
| *** Yanyanhu has joined #senlin | 01:50 | |
| *** sridhar_ram has quit IRC | 01:51 | |
| *** Liuqing has joined #senlin | 02:02 | |
| openstackgerrit | Qiming Teng proposed openstack/senlin: Add some backlog of release notes https://review.openstack.org/283909 | 02:30 |
|---|---|---|
| *** sridhar_ram has joined #senlin | 02:59 | |
| openstackgerrit | Qiming Teng proposed openstack/senlin: Revised TODO list https://review.openstack.org/283915 | 03:08 |
| openstackgerrit | Qiming Teng proposed openstack/senlin: Documentation for deletion policy https://review.openstack.org/283558 | 03:22 |
| *** zzxwill has joined #senlin | 03:25 | |
| zzxwill | Hello. Which Operating System is for the official install.sh and requirements.txt? CentOS/Redhat, Ubuntu? | 03:27 |
| Qiming | zzxwill, please consider install senlin using devstack way | 03:28 |
| Qiming | upstream packagers will help create RPM packages and DEB packages | 03:28 |
| Qiming | check doc here: http://docs.openstack.org/developer/senlin/install.html | 03:29 |
| *** sridhar_ram has quit IRC | 03:33 | |
| *** zzxwill has quit IRC | 03:48 | |
| *** Liuqing has quit IRC | 04:37 | |
| *** Liuqing has joined #senlin | 04:38 | |
| *** zzxwill has joined #senlin | 04:44 | |
| zzxwill | @Qiming, I will try. Thanks. I asked the question as I found a tiny issue when installing packages from requirements.txt https://bugs.launchpad.net/senlin/+bug/1549082. | 04:52 |
| openstack | Launchpad bug 1549082 in senlin "fatal error: Python.h: No such file or directory when installing applications from requirements" [Undecided,New] - Assigned to zzxwill (zzxwill) | 04:52 |
| zzxwill | For CentOS 7, python-dev needs to be installed before that. | 04:52 |
| Qiming | thanks, zzxwill, for letting us know | 05:19 |
| *** openstack has joined #senlin | 13:23 | |
| lixinhui | maybe still some principle about period task like the questions I asked at the beginning | 13:24 |
| lixinhui | Since registry is a table | 13:25 |
| lixinhui | I do not think we need to search the table from different places | 13:26 |
| lixinhui | so when you agree to healthManager, I think we can search there | 13:26 |
| lixinhui | but hp behave as a filter | 13:26 |
| lixinhui | now seems you mentioned a method, just let hp to do the search then checl/recover | 13:27 |
| lixinhui | where to synergy the health policy and healthManager | 13:27 |
| lixinhui | besides the conf of periodical interval | 13:28 |
| lixinhui | ? | 13:28 |
| Qiming | what if a service engine dies? | 13:29 |
| lixinhui | I can see the good design on parallel check/recover but did not get the key | 13:29 |
| Qiming | you will need logic to reconstruct the registry? | 13:29 |
| Qiming | we will allow for more than one engine service instances to run in parallel | 13:30 |
| Qiming | if each health manager instance maintain their own 'registry' in memory, how can we ensure a cluster is not checked by more than one health manager instance? | 13:31 |
| lixinhui | but different policy object search the same table is the bottelneck | 13:32 |
| Qiming | it is just a DB SQL | 13:32 |
| Qiming | by parallelization, we are not adding new SQL queries | 13:32 |
| Qiming | a health manager can keep a 'cache' of the clusters registered to it | 13:33 |
| lixinhui | you mean each policy will ask maitain a memory list | 13:33 |
| Qiming | just like what do have done in cluster.rt['nodes'] | 13:33 |
| lixinhui | instead of seach each period | 13:33 |
| lixinhui | oh | 13:34 |
| lixinhui | I mean the registry list | 13:34 |
| lixinhui | not the content of each cluster | 13:34 |
| Qiming | sigh, I'm talking about a metaphore | 13:34 |
| Qiming | we can do healthmanager.rt['clusters'] for caching purpose, right? | 13:35 |
| lixinhui | why? | 13:35 |
| Qiming | to save DB queries | 13:35 |
| lixinhui | since health manager will not do the cluster traverse any more | 13:35 |
| Qiming | health manager do cluster level iteration | 13:36 |
| Qiming | then create thread to run health policy's check/recover method | 13:36 |
| Qiming | a health policy's check/recover method is passed in with a cluster_id, then it does what the policy has defined to do | 13:37 |
| Qiming | a health policy's attach method does two things: 1. create a record into DB table, recording that this cluster's health is being managed by a given health manager; 2. add the cluster ID to the health manager's rt dictionary (cache) | 13:38 |
| Qiming | when health manager restarts, it reconstructs the rt dictionary from the DB table | 13:39 |
| Qiming | if we choose to make things a little bit simpler, we can ignore the caching part | 13:39 |
| lixinhui | seems got it | 13:42 |
| lixinhui | I will try to write some test firstly, then come to you for suggestions/confirmation | 13:43 |
| Qiming | with a diagram, you will see the whole picture | 13:43 |
| Qiming | the health_manager should have its own thread pool | 13:43 |
| Qiming | just like the scheduler | 13:43 |
| lixinhui | enen | 13:44 |
| lixinhui | no others :) I am waiting... | 13:47 |
| lixinhui | I am waiting for you to draw all the diagram | 13:48 |
| lixinhui | no other parts? | 13:48 |
| lixinhui | in the diagram | 13:48 |
| Qiming | ....... | 13:50 |
| Qiming | ignore the diagram part please | 13:51 |
| *** zzxwill has joined #senlin | 13:51 | |
| lixinhui | okay ... | 13:52 |
| Qiming | code is the best architecture, sometimes | 13:53 |
| lixinhui | so we can remove the tag of periodical_task from health_check | 13:55 |
| Qiming | yes I think so | 13:56 |
| lixinhui | since it running is trigged by healthmanager | 13:56 |
| lixinhui | ok | 13:56 |
| Qiming | yes, but I need to revisit the periodic_tasks code | 13:56 |
| Qiming | make sure I'm understanding them correctly | 13:56 |
| lixinhui | ok | 13:57 |
| Qiming | seems the default interval of periodic_tasks decorated methods is 60 seconds | 14:00 |
| lixinhui | en | 14:00 |
| lixinhui | but according to my initial testing, the tagged health_check will not run | 14:00 |
| Qiming | it is not registered | 14:01 |
| lixinhui | how to registry? | 14:01 |
| lixinhui | could you help to send me the link or config related? | 14:02 |
| Qiming | that decorator only means that method is kind of a periodic task | 14:02 |
| lixinhui | seems so | 14:02 |
| lixinhui | but no manager to call it up | 14:02 |
| Qiming | yes | 14:04 |
| Qiming | the 'periodic_tasks() method in Health_Manager class is triggered via a timer | 14:08 |
| lixinhui | yes | 14:11 |
| lixinhui | it runs well | 14:11 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OSC plugin for openstack cluster show https://review.openstack.org/284157 | 14:29 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OpenstackClient plugin for cluster policy create https://review.openstack.org/282818 | 14:49 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OpenstackClient plugin for cluster policy update https://review.openstack.org/282819 | 14:50 |
| *** lixinhui has quit IRC | 14:50 | |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OpenstackClient plugin for cluster policy delete https://review.openstack.org/282820 | 14:51 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: OSC plugin for openstack cluster policy type list https://review.openstack.org/282865 | 14:53 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OSC plugin for cluster policy type show https://review.openstack.org/282866 | 14:53 |
| Qiming | oh, my | 14:54 |
| openstackgerrit | Merged openstack/python-senlinclient: Add OpenstackClient plugin for cluster policy show https://review.openstack.org/282817 | 14:55 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OSC plugin for openstack cluster show https://review.openstack.org/284157 | 15:03 |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OSC plugin for openstack cluster create https://review.openstack.org/284184 | 15:16 |
| *** zzxwill has quit IRC | 15:42 | |
| *** zzxwill has joined #senlin | 15:44 | |
| *** zzxwill has quit IRC | 15:49 | |
| *** zzxwill has joined #senlin | 15:51 | |
| *** zzxwill has quit IRC | 15:55 | |
| *** Qiming has quit IRC | 16:00 | |
| openstackgerrit | Di XiaoLi proposed openstack/python-senlinclient: Add OSC plugin for openstack cluster update https://review.openstack.org/284221 | 16:05 |
| *** sridhar_ram has joined #senlin | 16:49 | |
| *** zzxwill has joined #senlin | 16:52 | |
| *** zzxwill has quit IRC | 16:56 | |
| openstackgerrit | Ayush Garg proposed openstack/senlin: Remove extra parameters from actions document https://review.openstack.org/284326 | 18:08 |
| *** zzxwill has joined #senlin | 18:52 | |
| *** zzxwill has quit IRC | 18:57 | |
| *** zzxwill has joined #senlin | 20:53 | |
| *** zzxwill has quit IRC | 20:59 | |
| *** sridhar_ram1 has joined #senlin | 21:32 | |
| *** sridhar_ram has quit IRC | 21:34 | |
| *** sridhar_ram1 has quit IRC | 22:36 | |
| *** sridhar_ram has joined #senlin | 22:38 | |
| *** zzxwill has joined #senlin | 22:55 | |
| *** zzxwill has quit IRC | 23:00 | |
| *** Qiming has joined #senlin | 23:15 | |
| *** sridhar_ram has quit IRC | 23:20 | |
| openstackgerrit | Qiming Teng proposed openstack/senlin: Refactor health manager for future work https://review.openstack.org/284193 | 23:31 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!