| *** zhurong has joined #senlin | 01:33 | |
| *** zhurong has quit IRC | 01:44 | |
| *** LHB has joined #senlin | 02:06 | |
| openstackgerrit | Lingyong Xu proposed openstack/senlin master: Replace oslo_utils.timeutils.isotime https://review.openstack.org/469790 | 02:22 |
|---|---|---|
| Qiming | hi | 02:29 |
| Qiming | LHB, online? | 02:29 |
| LHB | Qiming, Hi | 02:29 |
| Qiming | I was looking at the several blueprints you proposed | 02:30 |
| Qiming | thanks for leading those efforts | 02:30 |
| Qiming | when looking at the text description, I'm still not very clear about some of the details | 02:31 |
| Qiming | for example: https://blueprints.launchpad.net/senlin/+spec/cluster-scale-out-use-existing-node | 02:31 |
| Qiming | this one proposes scaling out a cluster to some existing nodes | 02:31 |
| Qiming | and https://blueprints.launchpad.net/senlin/+spec/add-attribute-fast-scaling-to-cluster | 02:31 |
| Qiming | proposes a similar function, scaling a cluster using nodes backed by a standby cluster | 02:32 |
| Qiming | so ... I'm wondering if we can merge these two proposals | 02:32 |
| *** zhurong has joined #senlin | 02:32 | |
| Qiming | anyway, a user has to specify the candicate nodes for a scale-out operation, if we follow the design in the first BP | 02:33 |
| LHB | I think scale out use existing nodes means if we have not create a standby cluster, but if there have any nodes created by the same profile with the cluster, maybe we could consider to use these nodes. | 02:37 |
| Qiming | then you can use cluster-node-add for the same purpose, right? | 02:38 |
| LHB | yes, cluster-node-add is manual, I want automatic to use scale-out to add these nodes . | 02:40 |
| Qiming | scale-out is not repeatable ... | 02:40 |
| Qiming | there are two options, based on my understandings | 02:41 |
| Qiming | either you do "cluster-scale-out -c 1 --candidate <a-node-id>", or you let senlin engine choose a node for you | 02:41 |
| Qiming | the first command is the same as cluster-node-add, because it explicitly specifies the node to add | 02:42 |
| Qiming | while the second option is very dangerous, senlin doesn't know what other nodes are used for | 02:42 |
| LHB | En...I thought to use second option, but have not considered the security. | 02:47 |
| Qiming | there must be some other constraints when adding a node to a cluster | 02:47 |
| Qiming | we cannot make such a decision simply based on profile type | 02:47 |
| Qiming | so ... from a user's perspective, your second BP makes better sense | 02:48 |
| openstackgerrit | Tuan Luong-Anh proposed openstack/senlin master: Replace assertTrue(isinstance()) with assertIsInstance() https://review.openstack.org/470817 | 02:49 |
| Qiming | previously, I have been thinking about adding a different property to 'Cluster' which specifies the shadow/standby cluster directly | 02:49 |
| Qiming | maybe this idea should be merged into your design as well | 02:49 |
| Qiming | since it is gonna be a very useful feature, I'd suggest we start with a more detailed design to make sure everyone understands where we are going | 02:50 |
| Qiming | what I'm actually suggesting is a spec file proposed to the senlin/doc/specs directory | 02:51 |
| Qiming | we can exchange thoughts using gerrit review process, which is much more convenient than the launchpad blueprint | 02:52 |
| Qiming | here is a template: http://git.openstack.org/cgit/openstack/senlin/tree/doc/specs/template.rst | 02:52 |
| LHB | Thanks for your suggestions, I will think seriously and then wirte a specs later, because I'm basy for someting about graduation recently. | 02:56 |
| Qiming | okay, take your time | 02:56 |
| Qiming | some drafts would be very helpful for the team to discuss | 02:57 |
| openstackgerrit | RUIJIE YUAN proposed openstack/senlin master: reduce DB interaction in affinity-policy https://review.openstack.org/470828 | 03:25 |
| *** ruijie_ is now known as ruijie | 03:26 | |
| *** XueFeng has joined #senlin | 03:59 | |
| *** XueFeng has quit IRC | 05:38 | |
| *** XueFeng has joined #senlin | 05:38 | |
| openstackgerrit | Merged openstack/senlin master: Remove usage of parameter enforce_type https://review.openstack.org/469717 | 06:15 |
| openstackgerrit | Qiming Teng proposed openstack/senlin master: Remove a TODO item https://review.openstack.org/470853 | 06:18 |
| chenyb4 | Hi, I have a question. How can i join senlin team ? | 06:18 |
| Qiming | hi, chenyb4, what do you mean by joining the team | 06:19 |
| chenyb4 | the nova allow user use this url https://launchpad.net/~nova join | 06:19 |
| Qiming | everyone working on the development is considered part of the team | 06:19 |
| Qiming | the launchpad team is meaningless, afaict | 06:20 |
| chenyb4 | ok, thanks | 06:20 |
| openstackgerrit | Qiming Teng proposed openstack/senlin master: Use get_rpc_transport instead of get_transport https://review.openstack.org/470855 | 06:27 |
| *** XueFeng has quit IRC | 06:28 | |
| *** XueFeng has joined #senlin | 06:29 | |
| *** elynn has joined #senlin | 06:29 | |
| *** LHB has quit IRC | 06:50 | |
| *** LHB has joined #senlin | 06:50 | |
| ruijie | hi Qiming, there? | 06:51 |
| Qiming | yes | 06:52 |
| ruijie | I am trying to use action's runtime data in policies instead of querying it from DB | 06:52 |
| ruijie | is that acceptable :) | 06:52 |
| openstackgerrit | Qiming Teng proposed openstack/senlin master: Remove a TODO item related to oslo.messaging https://review.openstack.org/470859 | 06:54 |
| Qiming | it is a little bit dangerous, but doable | 06:54 |
| Qiming | probably you will need a review of the Cluster (or Node) object initialization logic at the base Action class | 06:55 |
| Qiming | we need to ensure (at least) when an action is intialized, we are seeing the latest resource status from database. | 06:55 |
| Qiming | during Action execution, most of the time, the 'entity' is locked, so its properties won't get changed by other concurrent Actions | 06:56 |
| Qiming | that was the reason we 'invented' the runtime data thing | 06:56 |
| openstackgerrit | Merged openstack/senlin master: reduce DB interaction in affinity-policy https://review.openstack.org/470828 | 07:00 |
| ruijie | yes Qiming, that should be fine: http://git.openstack.org/cgit/openstack/senlin/tree/senlin/engine/actions/base.py#n467 | 07:01 |
| ruijie | we will load the action every time we process it, the action will load the entity then | 07:02 |
| Qiming | well, I meant this line: http://git.openstack.org/cgit/openstack/senlin/tree/senlin/engine/actions/cluster_action.py#n53 | 07:03 |
| Qiming | the entity is loaded afresh, that should be fine | 07:03 |
| ruijie | yes Qiming :) | 07:05 |
| *** elynn has quit IRC | 07:14 | |
| *** zhurong has quit IRC | 07:31 | |
| *** zhurong has joined #senlin | 07:40 | |
| openstackgerrit | RUIJIE YUAN proposed openstack/senlin master: revise deletion-policy to use action's runtime data https://review.openstack.org/470867 | 07:41 |
| *** zhurong has quit IRC | 08:20 | |
| *** zhurong has joined #senlin | 08:50 | |
| *** XueFeng has quit IRC | 09:30 | |
| *** XueFeng has joined #senlin | 09:30 | |
| *** zhurong has quit IRC | 09:59 | |
| *** zhurong has joined #senlin | 10:13 | |
| *** LHB has quit IRC | 10:28 | |
| *** zhurong has quit IRC | 12:04 | |
| *** catintheroof has joined #senlin | 12:35 | |
| openstackgerrit | Merged openstack/senlin master: Remove a TODO item related to oslo.messaging https://review.openstack.org/470859 | 13:17 |
| *** chenyb4_ has joined #senlin | 13:28 | |
| *** zhurong has joined #senlin | 13:46 | |
| openstackgerrit | Merged openstack/senlin master: Use get_rpc_transport instead of get_transport https://review.openstack.org/470855 | 14:40 |
| *** zhurong has quit IRC | 15:15 | |
| *** XueFeng has quit IRC | 15:40 | |
| *** XueFeng has joined #senlin | 15:41 | |
| *** chenyb4_ has quit IRC | 16:39 | |
| *** catintheroof has quit IRC | 21:53 | |
| *** catintheroof has joined #senlin | 23:13 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!