| Qiming | dtruong, hi, we haven't given that a deep thought yet. we did received some requests on it. | 01:19 |
|---|---|---|
| Qiming | our current suggestion for those use cases is to have the users develop downstream event processors | 01:19 |
| Qiming | when cluster scale is changed (or something similar happens), senlin emits event notifications. users can subscribe to those events for further processing. | 01:20 |
| Qiming | down this line is another topic we have just discussed during team meetup -- | 01:21 |
| Qiming | we may add a HTTP POST call as a driver for event generation | 01:21 |
| Qiming | we believe that is somehow equivalent to the lifecycle hooks you mentioned | 01:22 |
| *** openstackgerrit has joined #senlin | 03:13 | |
| *** ChanServ sets mode: +v openstackgerrit | 03:13 | |
| openstackgerrit | Merged openstack/senlin master: Clean useless decorator in test https://review.openstack.org/516540 | 03:13 |
| openstackgerrit | Merged openstack/senlin master: Remove override get_schema https://review.openstack.org/516230 | 03:24 |
| *** dtruong_ has joined #senlin | 03:43 | |
| dtruong_ | hi qiming. thanks for the answer. the problem with the downstream event processing is that it sends the notification after senlin already performed the scale in action and deleted the nodes. | 03:48 |
| dtruong_ | the goal of the lifecycle hooks is to pause the scale in action and allow another service or application to perform clean up on the nodes that were identified to be deleted. | 03:49 |
| dtruong_ | i'm working on an implementation that would another API to senlin for a scale in with lifecycle hook. | 03:50 |
| Qiming | em, that IS a problem. | 03:50 |
| Qiming | we have to tell which API request is the original one and which is the "confirmed" one | 03:50 |
| Qiming | sounds like we still need an out-of-band flag for differentiation | 03:51 |
| Qiming | however, the team did discussed this use case during our meetup last week | 03:52 |
| Qiming | the solution we came up is something like this: | 03:52 |
| Qiming | you can attach a deletion policy to your cluster, telling Senlin that each node should be granted a graceful period before being actually removed | 03:53 |
| Qiming | this is already supported | 03:53 |
| Qiming | what is missing is that Senlin should somehow send an alarm to your node (or some software) that a node is being destoryed | 03:54 |
| Qiming | that logic is missing, and we felt a need to support a generic HTTP sink for event notification | 03:55 |
| Qiming | you can do whatever logic during the time you get a notification and the time your node is deleted | 03:55 |
| Qiming | you can adjust the grace period if you have a lot of things to do during that interval | 03:56 |
| dtruong_ | i see. the one disadvantage i see is that we cannot notify senlin that the clean up has completed. in AWS there is a concept of lifecycle completion where the application can notify the scaling group that the clean up is done. | 04:01 |
| dtruong_ | in this case we would have to wait until the grace period expires. | 04:01 |
| Qiming | yep | 04:01 |
| dtruong_ | i guess my general question is how does the team feel about adding the equivalent of the AWS lifecycle hooks in senlin | 04:02 |
| dtruong_ | i was thinking of a design where is a new API for 'scale in with lifecycle hook'. | 04:04 |
| dtruong_ | that API would be similar to the existing 'scale in' API but also take a parameter for a zaqar queue. then during execution of that API, it would identify nodes for deletion same as the 'scale-in' API | 04:06 |
| dtruong_ | but instead of actually deleting the nodes it would send a notification to the zaqar queue of the nodes to be deleted. | 04:06 |
| dtruong_ | that would be the lifecycle hook for the application to perform the custom clean-up | 04:07 |
| *** AlexeyAbashkin has joined #senlin | 04:18 | |
| *** AlexeyAbashkin has quit IRC | 04:22 | |
| *** dtruong_ has quit IRC | 05:00 | |
| *** dtruong_ has joined #senlin | 05:06 | |
| *** maestropandy has joined #senlin | 05:31 | |
| *** maestropandy has quit IRC | 05:35 | |
| dtruong_ | qiming: i will submit a spec file for review that explains my idea in more detail. | 05:37 |
| Qiming | great! | 05:38 |
| *** dtruong_ has quit IRC | 05:39 | |
| *** maestropandy has joined #senlin | 06:03 | |
| *** maestropandy has left #senlin | 06:14 | |
| *** maestropandy has joined #senlin | 07:23 | |
| *** maestropandy1 has joined #senlin | 07:27 | |
| *** maestropandy has quit IRC | 07:29 | |
| *** maestropandy1 has left #senlin | 07:58 | |
| *** AlexeyAbashkin has joined #senlin | 07:58 | |
| *** liyi has joined #senlin | 08:02 | |
| *** liyi has quit IRC | 08:05 | |
| *** liyi has joined #senlin | 08:11 | |
| *** liyi has quit IRC | 08:22 | |
| *** liyi has joined #senlin | 08:30 | |
| *** maestropandy has joined #senlin | 09:38 | |
| *** maestropandy has quit IRC | 09:59 | |
| *** maestropandy has joined #senlin | 10:00 | |
| *** liyi has quit IRC | 10:07 | |
| *** liyi has joined #senlin | 10:07 | |
| *** liyi has quit IRC | 10:12 | |
| *** liyi has joined #senlin | 10:28 | |
| *** maestropandy has quit IRC | 10:29 | |
| *** liyi has quit IRC | 10:32 | |
| *** liyi has joined #senlin | 11:01 | |
| *** catintheroof has joined #senlin | 12:34 | |
| *** maestropandy has joined #senlin | 12:36 | |
| *** maestropandy has quit IRC | 12:38 | |
| *** liyi has quit IRC | 12:58 | |
| *** AlexeyAbashkin has quit IRC | 13:23 | |
| *** jmlowe_ has quit IRC | 13:36 | |
| *** AlexeyAbashkin has joined #senlin | 13:45 | |
| *** liyi has joined #senlin | 13:50 | |
| *** liyi has quit IRC | 13:54 | |
| *** jmlowe has joined #senlin | 14:04 | |
| *** jmlowe has quit IRC | 14:11 | |
| *** liyi has joined #senlin | 14:21 | |
| *** liyi has quit IRC | 14:25 | |
| *** jmlowe has joined #senlin | 14:27 | |
| *** AlexeyAbashkin has quit IRC | 15:53 | |
| *** liyi has joined #senlin | 16:18 | |
| *** liyi has quit IRC | 16:22 | |
| *** catintheroof has quit IRC | 16:26 | |
| *** catintheroof has joined #senlin | 16:30 | |
| *** dtruong has quit IRC | 16:45 | |
| *** dtruong has joined #senlin | 16:53 | |
| *** liyi has joined #senlin | 17:07 | |
| *** liyi has quit IRC | 17:12 | |
| *** AlexeyAbashkin has joined #senlin | 17:31 | |
| *** AlexeyAbashkin has quit IRC | 17:35 | |
| *** AlexeyAbashkin has joined #senlin | 18:09 | |
| *** AlexeyAbashkin has quit IRC | 18:14 | |
| *** liyi has joined #senlin | 19:16 | |
| *** liyi has quit IRC | 19:20 | |
| *** liyi has joined #senlin | 20:26 | |
| *** liyi has quit IRC | 20:30 | |
| *** jmlowe has quit IRC | 20:48 | |
| *** AlexeyAbashkin has joined #senlin | 21:08 | |
| *** AlexeyAbashkin has quit IRC | 21:13 | |
| *** jmlowe has joined #senlin | 21:41 | |
| *** catintheroof has quit IRC | 21:53 | |
| *** AlexeyAbashkin has joined #senlin | 22:08 | |
| *** AlexeyAbashkin has quit IRC | 22:13 | |
| *** liyi has joined #senlin | 22:45 | |
| *** liyi has quit IRC | 22:49 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!