| *** zhurong has joined #senlin | 00:44 | |
| *** shu-mutou-AWAY is now known as shu-mutou | 01:22 | |
| *** yanyanhu has joined #senlin | 01:33 | |
| openstackgerrit | Qiming Teng proposed openstack/senlin master: Pike-1 release notes https://review.openstack.org/456146 | 01:53 |
|---|---|---|
| openstackgerrit | Shan Guo proposed openstack/senlin master: Update DOC for the cluster update API https://review.openstack.org/456428 | 01:54 |
| *** guoshan has joined #senlin | 01:54 | |
| openstackgerrit | Shan Guo proposed openstack/senlin master: Update DOC for the cluster update API https://review.openstack.org/456428 | 01:56 |
| openstackgerrit | Merged openstack/python-senlinclient master: delete bash_completion in subcommand https://review.openstack.org/444617 | 01:58 |
| openstackgerrit | Merged openstack/python-senlinclient master: Updated from global requirements https://review.openstack.org/454535 | 01:58 |
| openstackgerrit | Merged openstack/python-senlinclient master: Fixed warnings when building docs https://review.openstack.org/434177 | 01:59 |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve check_object for health_policy_poll recover https://review.openstack.org/456187 | 02:03 |
| openstackgerrit | Merged openstack/senlin master: handle node which status is WARNING https://review.openstack.org/455542 | 02:13 |
| openstackgerrit | Merged openstack/senlin master: revise engine cluster obj to update runtime data https://review.openstack.org/456133 | 02:13 |
| openstackgerrit | Merged openstack/senlin master: Update DOC for the cluster update API https://review.openstack.org/456428 | 02:19 |
| openstackgerrit | Shu Muto proposed openstack/senlin-dashboard master: Refresh test environments https://review.openstack.org/456189 | 02:20 |
| *** ruijie has joined #senlin | 02:31 | |
| openstackgerrit | RUIJIE YUAN proposed openstack/senlin master: add health_check() to engine.cluster https://review.openstack.org/456432 | 02:55 |
| openstackgerrit | RUIJIE YUAN proposed openstack/senlin master: add health_check() to engine.cluster https://review.openstack.org/456432 | 03:06 |
| *** shu-mutou has quit IRC | 03:24 | |
| openstackgerrit | XueFeng Liu proposed openstack/senlin master: Revise the status_reason in node_check https://review.openstack.org/456438 | 03:29 |
| openstackgerrit | Merged openstack/senlin-dashboard master: Imported Translations from Zanata https://review.openstack.org/455535 | 03:33 |
| openstackgerrit | Merged openstack/senlin-dashboard master: Updated from global requirements https://review.openstack.org/456019 | 03:33 |
| openstackgerrit | Merged openstack/senlin-dashboard master: Refresh test environments https://review.openstack.org/456189 | 03:33 |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve do_check for health_policy_poll recover https://review.openstack.org/456187 | 03:39 |
| *** zhurong has quit IRC | 03:57 | |
| *** guoshan has quit IRC | 04:00 | |
| *** zhurong has joined #senlin | 04:16 | |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve do_check for health_policy_poll recover https://review.openstack.org/456187 | 04:24 |
| *** shu-mutou has joined #senlin | 04:27 | |
| openstackgerrit | XueFeng Liu proposed openstack/senlin master: Revise the status_reason in node_check https://review.openstack.org/456438 | 04:29 |
| openstackgerrit | XueFeng Liu proposed openstack/senlin master: Revise the status_reason in node_check https://review.openstack.org/456438 | 05:36 |
| *** yanyanhu has quit IRC | 05:46 | |
| *** yanyanhu has joined #senlin | 05:52 | |
| openstackgerrit | RUIJIE YUAN proposed openstack/senlin master: add health check to deletion policy https://review.openstack.org/456453 | 05:53 |
| *** ruijie has quit IRC | 05:57 | |
| *** dixiaoli has joined #senlin | 06:08 | |
| *** ruijie has joined #senlin | 06:13 | |
| xuhaiwei | hi XueFeng | 06:53 |
| openstackgerrit | Merged openstack/senlin master: Revise the status_reason in node_check https://review.openstack.org/456438 | 06:58 |
| Qiming | hi, ruijie, there? | 07:07 |
| ruijie | yes Qiming | 07:08 |
| Qiming | hi | 07:09 |
| Qiming | don't understand why we hack deletion policy to do node health check? | 07:09 |
| ruijie | em, if we have deletion policy bind to cluster | 07:09 |
| ruijie | we will choose the candidates in pre_op() | 07:10 |
| Qiming | yes | 07:10 |
| ruijie | so, as we discussed before, we should do health check before scaling | 07:10 |
| ruijie | then we can filter the bad nodes based on the health check | 07:10 |
| Qiming | right, but they are two different problems | 07:10 |
| Qiming | I'm seeing that we are mixing two things together again | 07:11 |
| Qiming | one of the design principle we had in senlin is to make policies orthogonal to each other | 07:11 |
| Qiming | if we are fixing the scaling behavior, we should actually consider the case when there is and there isn't a deletion policy | 07:12 |
| ruijie | em .. right Qiming, that is true. | 07:12 |
| Qiming | if we are tackling things this way, we are introducing a strong dependency from the scaling policy to the deletion policy | 07:12 |
| ruijie | so the plan now is add logic to deletion policy if exist, and add logic to do_xyz() too. | 07:12 |
| Qiming | wait ... | 07:13 |
| ruijie | :( | 07:13 |
| Qiming | if we are adding this to the deletion policy | 07:13 |
| Qiming | do you want to expose it as a policy option as well? | 07:14 |
| ruijie | see, the situations are: 1. scaling policy, no deletion policy, 2. no scaling policy, deletion policy, 3. have both, 4. both are not exist .. | 07:14 |
| Qiming | if we are adding 'check_health_first' into scaling policy, for example, there will be conflicts ? | 07:15 |
| Qiming | right | 07:15 |
| ruijie | scaling policy is high priority | 07:15 |
| ruijie | the health_check parameter will be added to scaling policy. | 07:15 |
| Qiming | okay, let's put the deletion policy thing aside | 07:15 |
| Qiming | focus on scaling policy now | 07:16 |
| ruijie | in scaling policy logic, will dump this parameter to action.data.deletion | 07:16 |
| ruijie | cause, the scaling policy actually only count the count .. | 07:16 |
| Qiming | let's take a step back from modifying the policy | 07:16 |
| Qiming | understood | 07:16 |
| ruijie | Qiming, you mean put all the policies aside? | 07:17 |
| Qiming | the thing is ... if we don't have a clear vision where we are heading, it is too easy to get things messed up | 07:17 |
| Qiming | I mean we focus on scaling policy for now, we can discuss the deletion policy later | 07:17 |
| ruijie | okay, Qiming, just like I said, the only thing we can do is dump the 'health_check' parameter to action.data | 07:18 |
| Qiming | during the discussion, I am also trying to gain a better understanding whether it makes sense to add this to a policy spec at all | 07:18 |
| Qiming | okay, for the scaling scenario | 07:19 |
| Qiming | do we have a plan to make the scaling operations consistent with and without a scaling policy? | 07:19 |
| ruijie | if we have one, we dump it to action.data, if we don't we try to get it from the request object | 07:20 |
| Qiming | actually, you may have more than one policy, right? | 07:20 |
| ruijie | yes Qiming | 07:20 |
| Qiming | here is a scenario | 07:20 |
| Qiming | I have a cluster that needs to be scaled down quickly if there are some events happened, it should be scaled a little bit slowly when another thing happened | 07:21 |
| Qiming | we actually allow many instances of scaling policies attached to the same cluster | 07:22 |
| Qiming | would it be a problem for the health checking behavior? | 07:23 |
| ruijie | only one scaling policy could be enabled at the same time ? | 07:24 |
| Qiming | no, more than one, we actually allow many scaling policies to be attached and enabled | 07:25 |
| Qiming | and ... expose scaling operation apis for "manual" scaling | 07:25 |
| ruijie | em .. let me have a quick check .. | 07:25 |
| Qiming | and cluster_resize ... | 07:25 |
| Qiming | this line: http://git.openstack.org/cgit/openstack/senlin/tree/senlin/policies/scaling_policy.py#n126 | 07:26 |
| ruijie | okay.. then that will be a problem. | 07:27 |
| Qiming | it may and may not be a problem | 07:27 |
| Qiming | I'm not sure | 07:27 |
| Qiming | just wanted to point out some possibilities | 07:28 |
| ruijie | for the current situation, I'd suggest we add the parameter to deletion policy. | 07:28 |
| Qiming | back to the scaling behavior where we want to do a health check | 07:28 |
| ruijie | since that is not very necessary to do health check for a scale out procedure | 07:28 |
| Qiming | we have two options | 07:28 |
| Qiming | 1) do this operation in action handler, 2) do this in policy implementation | 07:29 |
| ruijie | em.. but the deletion policy will chose the candidates to delete | 07:29 |
| Qiming | alright, forget deletion policy for the moment | 07:30 |
| Qiming | we can revisit that later | 07:30 |
| ruijie | okay | 07:30 |
| Qiming | the key requirement for this whole bp is to make sure when we are removing nodes from a cluster, we can do a smart job | 07:31 |
| ruijie | without the deletion policy, we could create number of node actions to do it | 07:31 |
| Qiming | which means we will remove the bad ones first | 07:31 |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve check_object for health_policy_poll recover https://review.openstack.org/456187 | 07:31 |
| Qiming | this requirement remains true no matter users attached any policy to a cluster or not, right? | 07:32 |
| ruijie | yes Qiming | 07:32 |
| Qiming | okay, so if I were you | 07:32 |
| Qiming | I will do this in a different order | 07:32 |
| Qiming | I'll make sure node health check can be invoked from relevant node actions, when specified | 07:33 |
| Qiming | if needed, we may want to add some parameters to the API requests | 07:33 |
| Qiming | then next thing I will check is for user's convenience | 07:34 |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve check_object for health_policy_poll recover https://review.openstack.org/456187 | 07:34 |
| Qiming | if they are lazy, they want this behavior to be configurable rather than provided as parameters each and every time | 07:34 |
| Qiming | this can be done by improving the relevant policies | 07:35 |
| Qiming | a policy can help decide whether node health check should be done, instead of performing the health check operation directly | 07:35 |
| Qiming | in other words, a policy impacts the behavior of actions, but they are better not doing such operations | 07:36 |
| Qiming | they write their decisions into action.data | 07:36 |
| Qiming | the only exception for this is the LB policy, which was doing more than just policy decisions, they create resources and manage them | 07:38 |
| ruijie | okay, QIming, you mean, we do it from API | 07:40 |
| *** yanyanhu has quit IRC | 07:40 | |
| *** yanyanhu has joined #senlin | 07:40 | |
| Qiming | em, as a general rule | 07:40 |
| ruijie | then if needed we may revisit the policy | 07:40 |
| Qiming | yes | 07:41 |
| Qiming | don't start from hacking policies | 07:41 |
| ruijie | so, without any policies now, we add a parameter 'health_check' to req obj | 07:41 |
| ruijie | in cluster_action layer, we get it from action.inputs to make decions | 07:42 |
| Qiming | yep | 07:42 |
| ruijie | and create node_check actions, or invoke the cluster.health_check() | 07:42 |
| Qiming | once these logic are good and stable | 07:42 |
| Qiming | we can pump them into policies | 07:43 |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve check_object for health_policy_poll recover https://review.openstack.org/456187 | 07:43 |
| ruijie | okay Qiming, that makes good sense | 07:44 |
| Qiming | we have done something similar for node-recover and cluster-recover | 07:45 |
| Qiming | ref: http://git.openstack.org/cgit/openstack/senlin/tree/senlin/api/openstack/history.rst#n78 | 07:45 |
| ruijie | yes Qiming, noticed that change. | 07:53 |
| ruijie | will think about the logic | 07:53 |
| -openstackstatus- NOTICE: zuul was restarted due to an unrecoverable disconnect from gerrit. If your change is missing a CI result and isn't listed in the pipelines on http://status.openstack.org/zuul/ , please recheck | 08:52 | |
| openstackgerrit | Merged openstack/senlin master: Pike-1 release notes https://review.openstack.org/456146 | 09:24 |
| *** shu-mutou is now known as shu-mutou-AWAY | 09:28 | |
| *** yanyanhu has quit IRC | 10:02 | |
| *** zhurong has quit IRC | 10:17 | |
| *** zhurong has joined #senlin | 10:45 | |
| openstackgerrit | yangyide proposed openstack/senlin master: Improve check_object for health_policy_poll recover https://review.openstack.org/456187 | 11:10 |
| *** dixiaoli has quit IRC | 11:31 | |
| *** guoshan has joined #senlin | 11:42 | |
| *** guoshan has quit IRC | 13:33 | |
| *** catintheroof has joined #senlin | 13:44 | |
| *** catintheroof has quit IRC | 13:45 | |
| *** catintheroof has joined #senlin | 13:45 | |
| *** guoshan has joined #senlin | 13:50 | |
| *** catintheroof has quit IRC | 13:50 | |
| *** catintheroof has joined #senlin | 13:51 | |
| *** zhurong has quit IRC | 13:57 | |
| *** catintheroof has quit IRC | 14:52 | |
| *** catintheroof has joined #senlin | 14:52 | |
| *** zigo_ has joined #senlin | 15:09 | |
| *** zigo has quit IRC | 15:10 | |
| *** zigo_ is now known as zigo | 15:12 | |
| *** zigo is now known as Guest52586 | 15:12 | |
| *** catintheroof has quit IRC | 15:27 | |
| *** catintheroof has joined #senlin | 15:27 | |
| *** catintheroof has quit IRC | 15:27 | |
| *** guoshan has quit IRC | 15:31 | |
| *** XueFeng has quit IRC | 15:32 | |
| openstackgerrit | XueFeng Liu proposed openstack/senlin master: Use 'nosetests -v' replace './run_tempest.sh -N' https://review.openstack.org/456677 | 15:43 |
| *** openstack has joined #senlin | 16:37 | |
| *** catintheroof has quit IRC | 16:44 | |
| -openstackstatus- NOTICE: The Gerrit service on http://review.openstack.org is being restarted to address hung remote replication tasks. | 19:53 | |
| *** zhenguo has quit IRC | 20:32 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!