| openstackgerrit | Merged openstack/smaug: Make Plan parameterss a DictOfDict https://review.openstack.org/309748 | 01:44 |
|---|---|---|
| openstackgerrit | Merged openstack/smaug: Fix resource name can not be inserted into db https://review.openstack.org/310128 | 01:44 |
| openstackgerrit | Merged openstack/smaug-dashboard: [Checkpoints] Submit index view function https://review.openstack.org/307572 | 02:06 |
| openstackgerrit | Merged openstack/smaug-dashboard: [Operation Logs] Submit index view function https://review.openstack.org/307729 | 02:06 |
| openstackgerrit | chenying proposed openstack/python-smaugclient: Fix creating scheduled_operations error https://review.openstack.org/311921 | 02:50 |
| *** zhangshuai has joined #openstack-smaug | 03:04 | |
| *** zhangshuai has quit IRC | 03:21 | |
| *** yuval has joined #openstack-smaug | 06:34 | |
| openstackgerrit | chenying proposed openstack/python-smaugclient: Fix creating scheduled_operations error https://review.openstack.org/311921 | 06:56 |
| *** wangfeng has joined #openstack-smaug | 07:46 | |
| *** c00281451 has joined #openstack-smaug | 07:47 | |
| *** c00281451 is now known as chenzeng | 07:50 | |
| chenzeng | yuval:good morning! how are you? | 07:50 |
| *** wangfeng has quit IRC | 08:27 | |
| yuval | chenzeng: hey, how are you? | 08:30 |
| chenzeng | yuval:good. long time to talk to you. | 08:41 |
| chenzeng | yuval:i have two issues to talk to you. 1. do you have to time to review the 3 patches, https://review.openstack.org/#/c/282263/,https://review.openstack.org/#/c/309898,https://review.openstack.org/#/c/310394. | 08:44 |
| chenzeng | i hope they can be merged ASAP. | 08:44 |
| yuval | ok, what is the second issue? | 08:47 |
| chenzeng | 2. I think the solution that send the creating scheduled operation request by call and others by cast is too weak. do we need to add a scheduler to chose the target node like the scheduler in nova? | 08:48 |
| yuval | what do you mean by "too weak"? | 08:49 |
| chenzeng | we use cast to broadcast the requests, so we can not make sure the requests are executed successfully. we just notify the operation engine service to do, but we can not treat the exceptions. | 08:55 |
| openstackgerrit | zengchen proposed openstack/smaug: Implement protect operation https://review.openstack.org/309898 | 08:56 |
| openstackgerrit | chenpengzi proposed openstack/smaug: Fix fullstack test_plan_list https://review.openstack.org/312008 | 10:15 |
| yuval | ping chenzeng | 10:38 |
| chenzeng | yuval:sorry, i went to have dinner. don't you have a rest? | 10:47 |
| yuval | chenzeng: we do, I just had lunch | 10:48 |
| yuval | chenzeng: (burger) | 10:48 |
| chenzeng | yuval:oh, your lunch is so simple. | 10:50 |
| chenzeng | yuval:how about that question? | 10:50 |
| chenzeng | yuval:what's your opinion? | 10:51 |
| yuval | chenzeng: I agree that there might be an issue here | 10:52 |
| chenzeng | yuval:so, we have to get a better solution. | 10:53 |
| yuval | chenzeng: have you looked in oslo.messaging? maybe there is a solution for that? maybe there is a planned solution? | 10:53 |
| chenzeng | yuval:no, i will look at it tomorrow. | 10:55 |
| yuval | chenzeng: how do new host register at nova's scheduler? | 10:55 |
| yuval | if we want a scheduler-like solution, doesn't it imply a central scheduler? | 10:56 |
| chenzeng | yuval:i think when the compute service start, it will register to scheduler. nova have a central scheduler service. | 10:58 |
| chenzeng | yuval:do we need to send a email of openstack to get the other guys's idea? | 11:01 |
| chenzeng | yuval:i think for the trigger, we use cast interface of RPC, but for the operation, it is better to add a scheduler to find a node to create/delete operations. | 11:04 |
| yuval | chenzeng: why not use scheduler for operation also? | 11:05 |
| yuval | *for trigger | 11:06 |
| chenzeng | because for all nodes, they have same all triggers. | 11:07 |
| yuval | but what if registering a trigger use 'cast()' fails due to an exception? | 11:11 |
| yuval | chenzeng: ? | 11:16 |
| chenzeng | yuval:yes | 11:16 |
| yuval | chenzeng: but what if registering a trigger use 'cast()' fails due to an exception? | 11:18 |
| chenzeng | the trigger are stored in the memory, there is not bad influence when the create/delete/update trigger failed. | 11:18 |
| yuval | chenzeng: what if the update trigger fails, and then we register an operation to that trigger? | 11:19 |
| yuval | chenzeng: it will use the configuration before the update | 11:19 |
| chenzeng | yuval:yes, you are right. it is a problem for the consistency. | 11:21 |
| chenzeng | yuval:but for trigger, we don't need the scheduler to get a node. the target of trigger's operation is all nodesl | 11:24 |
| yuval | chenzeng: yes. and if we use scheduler which does 'call()' on every service, it will need to iterate each one of them | 11:26 |
| yuval | chenzeng: all nodes = all services? | 11:26 |
| chenzeng | yuval: all nodes = all services? yes | 11:26 |
| chenzeng | yuval:is it too slow for 'call()' on every service if we have multiple services. | 11:28 |
| yuval | chenzeng: might very well be | 11:29 |
| yuval | chenzeng: I'm trying to find how other openstack projects deal with that | 11:30 |
| yuval | chenzeng: can you take a look at nova/scheduler/rpcapi.py ? | 11:33 |
| chenzeng | yuval:yes. | 11:34 |
| yuval | chenzeng: it uses fanout cast() without verifying and exceptions | 11:35 |
| chenzeng | yuval:so, you mean we can still use cast? | 11:36 |
| chenzeng | ingore the exceptions. | 11:36 |
| yuval | I think for now we can still use cast | 11:54 |
| yuval | but we should consider the case of what happens when we register an operation to a service, and the service goes down | 11:54 |
| chenzeng | yuval:ok. so we don't add a scheduler service, right. | 12:03 |
| yuval | chenzeng: what do you think about the above issue? (operation engine service goes down when it has operations registered to it) | 12:04 |
| chenzeng | yuval:we can restart the service which will register the operation again. | 12:06 |
| yuval | chenzeng: what if the service goes down and stays down? | 12:06 |
| chenzeng | yuval:we can start the service on another node and migrate the operations to this service. | 12:07 |
| yuval | chenzeng: but we need to detect that the service is down and the operation didn't start | 12:07 |
| chenzeng | sorry not quite understand. | 12:08 |
| chenzeng | what do you mean for ' the operation didn't start' | 12:09 |
| yuval | ok | 12:09 |
| yuval | say we have a trigger: 12:00 every day | 12:09 |
| chenzeng | yes | 12:09 |
| yuval | we register operation "protect project A" to this trigger | 12:09 |
| yuval | this register the operation in the database, and also sends a 'call()' to register the operation to the trigger on one service | 12:11 |
| yuval | now everything works fine for a week | 12:11 |
| chenzeng | ok | 12:11 |
| chenzeng | ok | 12:12 |
| chenzeng | i understand | 12:12 |
| yuval | after that, we take this specific operation engine service down, and let it stay down | 12:12 |
| yuval | operation never happens | 12:12 |
| chenzeng | you worry the operations can not be triggered on time for the service down | 12:12 |
| yuval | yep | 12:13 |
| yuval | this is an important issue | 12:14 |
| chenzeng | yes, i agree. | 12:14 |
| chenzeng | that is the next step we must do. | 12:15 |
| chenzeng | we can discuss in the meeting later. | 12:15 |
| chenzeng | yuval:sorry, i have to go. see you later. | 12:16 |
| yuval | chenzeng: bye | 12:16 |
| *** wangfeng has joined #openstack-smaug | 12:36 | |
| *** saggi has joined #openstack-smaug | 13:50 | |
| openstackgerrit | WeAreFormalGroup proposed openstack/smaug: Implement list checkpoints https://review.openstack.org/310710 | 13:52 |
| *** x00350071 is now known as xiangxinyong | 13:56 | |
| *** chenying_ has joined #openstack-smaug | 14:00 | |
| *** yinweimac has joined #openstack-smaug | 14:00 | |
| *** pzchen has joined #openstack-smaug | 14:03 | |
| *** pzchen has quit IRC | 14:15 | |
| *** pzchen has joined #openstack-smaug | 14:15 | |
| *** zhangshuai has joined #openstack-smaug | 14:24 | |
| *** pzchen has quit IRC | 14:27 | |
| *** gampel has joined #openstack-smaug | 14:29 | |
| *** Guest93630 has joined #openstack-smaug | 14:34 | |
| Guest93630 | guys,is there a meeting holding tonight? | 14:35 |
| xiangxinyong | yeah | 14:36 |
| xiangxinyong | on the #openstack-meeting | 14:36 |
| *** gampel has quit IRC | 14:39 | |
| *** gampel has joined #openstack-smaug | 14:41 | |
| *** zhangshuai2 has joined #openstack-smaug | 14:42 | |
| Guest93630 | xiangxinyong, xiexie! | 14:44 |
| *** Guest93630 is now known as NeoL | 14:49 | |
| *** zhangshuai2 has quit IRC | 14:50 | |
| *** NeoL is now known as Bourne_L | 14:50 | |
| *** yinweimac has quit IRC | 15:01 | |
| *** chenying_ has quit IRC | 15:01 | |
| *** Bourne_L has quit IRC | 15:02 | |
| *** zhangshuai has quit IRC | 15:02 | |
| *** zhangshuai_ has joined #openstack-smaug | 15:03 | |
| *** yuval has quit IRC | 15:06 | |
| *** zhangshuai_ has quit IRC | 15:09 | |
| *** saggi has quit IRC | 15:59 | |
| *** wangfeng has quit IRC | 16:17 | |
| *** gampel has quit IRC | 17:39 | |
| *** gampel has joined #openstack-smaug | 17:41 | |
| *** gampel has quit IRC | 17:49 | |
| *** saggi has joined #openstack-smaug | 19:02 | |
| *** zhonghua-lee has quit IRC | 19:42 | |
| *** zhonghua-lee has joined #openstack-smaug | 19:51 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!