| *** shubhams_ has joined #openstack-zun | 01:10 | |
| *** shubhams_ has quit IRC | 01:12 | |
| *** yanyanhu has joined #openstack-zun | 01:39 | |
| *** hongbin has joined #openstack-zun | 02:24 | |
| hongbin | yuanying: ping | 02:29 |
|---|---|---|
| yuanying | hi | 02:29 |
| hongbin | yuanying: You mentioned that you are a big fan to use etcd for passing message ? :) | 02:29 |
| yuanying | Not so fan ;) | 02:29 |
| hongbin | What is your idea? | 02:30 |
| *** manikanta_tadi has joined #openstack-zun | 02:30 | |
| yuanying | Only just I found that `taskflow` uses kvs for message passing | 02:31 |
| hongbin | I saw they have a zookeeper backend | 02:31 |
| yuanying | yes | 02:31 |
| hongbin | But it seems they used it as data store | 02:31 |
| yuanying | Oh really? | 02:31 |
| hongbin | I couldn't find how they pass msg in zookeeper | 02:32 |
| hongbin | BTW, I have summarized the possible solutions I knew: https://etherpad.openstack.org/p/zun-container-state-management | 02:34 |
| hongbin | yuanying: ^^ | 02:34 |
| yuanying | ok | 02:35 |
| yuanying | I'll check it | 02:35 |
| hongbin | yeah, will discuss that in the team meeting | 02:36 |
| yuanying | Ah, I'll not be able to join team meeting this week, due to our company's meeting | 02:37 |
| hongbin | yuanying: NP | 02:37 |
| hongbin | yuanying: Then, if you have a chance, appreciate your comment in the etherpad :) | 02:37 |
| yuanying | hongbin: yes, of course | 02:38 |
| hongbin | thx! | 02:38 |
| yanyanhu | hi, hongbin, still around? | 02:40 |
| *** mkrai has joined #openstack-zun | 02:40 | |
| hongbin | yanyanhu: yes | 02:40 |
| yanyanhu | hi, about tooz | 02:40 |
| yanyanhu | I think I misunderstood its gola | 02:41 |
| yanyanhu | s/gola/goal | 02:41 |
| hongbin | that is fine | 02:41 |
| yanyanhu | it's for DLM not common data storage | 02:41 |
| hongbin | yes, looks like it is | 02:41 |
| yanyanhu | so it's not suitable for container state management :) | 02:41 |
| hongbin | yes | 02:41 |
| yanyanhu | we can leverage it to support service failure detection or HA | 02:41 |
| hongbin | how? | 02:42 |
| yanyanhu | e.g. we can use it to manage the status of each zun-conductor or zun-api | 02:42 |
| yanyanhu | and perform failover when failure happened | 02:42 |
| *** shubhams_ has joined #openstack-zun | 02:43 | |
| yanyanhu | kinda heartbeat mechanism | 02:43 |
| yanyanhu | this is one possible use case of tooz | 02:43 |
| yanyanhu | and also lock management actually | 02:43 |
| hongbin | interesting idea | 02:43 |
| yanyanhu | for some critical resources that could be accessed concurrently by multiple agents | 02:43 |
| yanyanhu | yes, I think we can have some try in future :) | 02:43 |
| hongbin | The lock management is possibly useful | 02:44 |
| yanyanhu | yes, I think heat team has made some try to use it for stack lock management | 02:44 |
| yanyanhu | can borrow some experience from them if possible | 02:45 |
| hongbin | sure | 02:45 |
| yanyanhu | anyway, for data storage, looks like etcd is better choice :) | 02:45 |
| hongbin | hm | 02:45 |
| hongbin | if etcd is the right choice, then we can use the kubernetes architecture | 02:46 |
| hongbin | architecture 1: https://etherpad.openstack.org/p/zun-container-state-management | 02:46 |
| yanyanhu | hongbin, yes, as you mentioned both authentication and transaction are not well supported | 02:47 |
| hongbin | yes | 02:48 |
| yanyanhu | but I feel authentication could not be a big problem for access to etcd will only come from zun service | 02:48 |
| hongbin | yes | 02:48 |
| yanyanhu | actually mysql has the same problem if the user credential is leaked :) | 02:48 |
| yanyanhu | transaction could be a problem for concurrency issue | 02:49 |
| hongbin | Maybe the solution is to make them listen to localhost | 02:49 |
| yanyanhu | hongbin, yes, that's possible | 02:49 |
| yanyanhu | hi, will leave for a while | 02:49 |
| yanyanhu | back soon | 02:49 |
| hongbin | k | 02:49 |
| hongbin | mkrai: hey | 02:50 |
| hongbin | mkrai: I tried to find you last week, but couldn't catch you | 02:50 |
| mkrai | Hi hongbin | 02:50 |
| mkrai | Ohh on IRC? | 02:51 |
| hongbin | mkrai: yes, but couldn't find you online | 02:51 |
| mkrai | Our timezone is completely different :) | 02:51 |
| *** shubhams_ has quit IRC | 02:51 | |
| hongbin | mkrai: I tried to find you because you tried to ping me in before | 02:51 |
| hongbin | mkrai: sup | 02:51 |
| mkrai | Yes I wanted to discuss about the conductor part | 02:52 |
| hongbin | yes | 02:52 |
| mkrai | Whether we really want to have it or not? | 02:52 |
| hongbin | I am not sure | 02:52 |
| mkrai | We can discuss it in today's open discussion | 02:52 |
| hongbin | Depends on the overall architecture | 02:52 |
| hongbin | sure | 02:53 |
| hongbin | mkrai: I could tell you what I think | 02:53 |
| mkrai | Sure | 02:53 |
| hongbin | mkrai: conductor is useful if there is a DB | 02:53 |
| hongbin | THen, conductor is a proxy to the DB | 02:53 |
| hongbin | Normally, for DB schema upgrade | 02:53 |
| mkrai | Yes and we don't have db, so this use case stands null | 02:54 |
| hongbin | Seems so | 02:54 |
| mkrai | Similarly we can take out its usecases and see whether they actually fit in our architecture or not | 02:54 |
| *** shubhams_ has joined #openstack-zun | 02:54 | |
| hongbin | If we choose not to use db | 02:55 |
| mkrai | I saw nova's conductor code and its seems it is only deliberating calls to another services and db part | 02:56 |
| mkrai | Talk to scheduler, then talk to compute and the pass response to api | 02:56 |
| hongbin | Yes, nova conductor is more than a proxy | 02:56 |
| hongbin | It contains business logic | 02:57 |
| mkrai | Yes I agree | 02:57 |
| hongbin | OK. Team meeting | 03:00 |
| hongbin | mkrai: Qiming flwang1 flwang : Hey, team meeting if you have a chance :) | 03:01 |
| *** klkumar has joined #openstack-zun | 03:24 | |
| *** mkrai has quit IRC | 03:26 | |
| *** mkrai has joined #openstack-zun | 03:27 | |
| *** manikanta_tadi has quit IRC | 03:51 | |
| *** sudipto has joined #openstack-zun | 04:00 | |
| *** shubhams_ has left #openstack-zun | 04:00 | |
| *** hongbin has quit IRC | 04:00 | |
| sudipto | About the pluggable scheduler thingy - hongbin - we will maybe catch up at a saner time for you. | 04:00 |
| mkrai | yanyanhu, I was asking whether the nova-scheduler can fit in for containers also? | 04:01 |
| mkrai | sudipto, ^ | 04:02 |
| yanyanhu | mkrai, I see. I think that depends on how we allocate container resource per user's requirement | 04:02 |
| *** Namrata has joined #openstack-zun | 04:02 | |
| sudipto | mkrai, that's a very good point. | 04:02 |
| yanyanhu | if the way user requests for container is quite different from VM, it will be difficult to reuse nova scheduler | 04:02 |
| Namrata | yeah true that | 04:02 |
| mkrai | Shceduler doesn't do anything specific to vm IMO | 04:02 |
| sudipto | mkrai, are we looking at using "nova boot" to boot a container? | 04:03 |
| yanyanhu | mkrai, right, that's my point | 04:03 |
| mkrai | It picks up host based on user requirement like the host resource, zones etc | 04:03 |
| mkrai | And that all applies to containers also | 04:03 |
| yanyanhu | +1 | 04:03 |
| sudipto | it does - something the CPUFilter for instance. Eventually it depends on how your runtime behaves. | 04:03 |
| mkrai | I am not sure about all the filters | 04:04 |
| sudipto | Do you want to allocate a certain 'cpu share' with your container following the vcpu based model? | 04:04 |
| mkrai | But I really want to see some that doesn't fit for us | 04:04 |
| sudipto | it sounds like - it may work - if we recommend our own set of filters to use. | 04:04 |
| yanyanhu | I guess we just need different pluggable scheduling filter? | 04:05 |
| sudipto | yanyanhu, +1 | 04:05 |
| yanyanhu | to handle those variation | 04:05 |
| mkrai | Yes agree | 04:05 |
| mkrai | Different filters for containers | 04:05 |
| sudipto | it sounds like we may have to write additional filters there. | 04:05 |
| mkrai | sudipto, +1 | 04:06 |
| yanyanhu | yes, have the same feeling | 04:06 |
| sudipto | and discard some of the existing ones that don't apply | 04:06 |
| mkrai | But we can't use nova scheduler at moment | 04:06 |
| yanyanhu | hi, guys, I have to leave, ttyl. Will read the log to catch your discussion later. | 04:06 |
| mkrai | yanyanhu, Bye! | 04:06 |
| sudipto | mkrai, i don't think this code path - and the one we discussed with respect to zun specific runtime - can be completely independent. | 04:06 |
| yanyanhu | bye | 04:06 |
| sudipto | yanyanhu, bbye | 04:07 |
| sudipto | sorry i mean - they should be kept independent | 04:07 |
| sudipto | that way - if we struggle with the nova integration - atleast the zun runtime path will be completely within our own control to implement. | 04:07 |
| mkrai | sudipto, Agree and we should have our own scheduler | 04:08 |
| Namrata | agreed | 04:08 |
| sudipto | mkrai, yeah a reference scheduler may be. I would go for a very pluggable scheduler with the approach of replacing it with mesos at any point in time. | 04:08 |
| mkrai | sudipto, I will be right back. Have team meeting | 04:08 |
| mkrai | Sorry | 04:08 |
| sudipto | alrite, i will catch up during the day! See ya later1 | 04:08 |
| mkrai | Please be online. I will catch you later today :) | 04:09 |
| mkrai | Bye sudipto Namrata | 04:09 |
| Namrata | i will also join | 04:09 |
| Namrata | bye | 04:09 |
| *** yanyanhu has quit IRC | 04:20 | |
| *** yanyanhu has joined #openstack-zun | 04:49 | |
| *** Namrata has quit IRC | 04:50 | |
| *** sudipto has quit IRC | 04:57 | |
| *** chandankumar has joined #openstack-zun | 04:58 | |
| *** flwang1 has quit IRC | 05:12 | |
| *** janki has joined #openstack-zun | 05:22 | |
| *** janki has quit IRC | 05:27 | |
| *** janki has joined #openstack-zun | 05:28 | |
| *** mkrai has quit IRC | 05:37 | |
| *** Namrata has joined #openstack-zun | 05:41 | |
| *** klkumar has quit IRC | 05:45 | |
| *** mkrai has joined #openstack-zun | 05:50 | |
| *** flwang1 has joined #openstack-zun | 06:03 | |
| *** mkrai has quit IRC | 06:14 | |
| *** mikelk has joined #openstack-zun | 06:52 | |
| *** mkrai has joined #openstack-zun | 07:01 | |
| *** sudipto has joined #openstack-zun | 07:07 | |
| *** sudipto has quit IRC | 07:18 | |
| *** sudipto has joined #openstack-zun | 07:27 | |
| *** sudipto has quit IRC | 07:47 | |
| *** sudipto has joined #openstack-zun | 08:02 | |
| *** mkrai has quit IRC | 08:04 | |
| *** mikelk has quit IRC | 08:09 | |
| *** mikelk has joined #openstack-zun | 08:13 | |
| *** mkrai has joined #openstack-zun | 08:20 | |
| *** sudipto has quit IRC | 08:45 | |
| *** yanyanhu has quit IRC | 10:23 | |
| *** itzdilip has quit IRC | 11:46 | |
| *** mkrai has quit IRC | 11:55 | |
| *** yanyanhu has joined #openstack-zun | 12:54 | |
| *** chandankumar has quit IRC | 13:10 | |
| *** yanyanhu has quit IRC | 13:36 | |
| *** yanyanhu has joined #openstack-zun | 13:37 | |
| *** sudipto_ has joined #openstack-zun | 13:46 | |
| *** sudipto has joined #openstack-zun | 13:46 | |
| *** sudipto_ has quit IRC | 13:47 | |
| *** sudipto has quit IRC | 13:47 | |
| *** sudipto has joined #openstack-zun | 13:47 | |
| *** sudipto has quit IRC | 13:48 | |
| *** sudipto has joined #openstack-zun | 13:48 | |
| *** janki has quit IRC | 14:04 | |
| *** yanyanhu has quit IRC | 14:04 | |
| *** hongbin has joined #openstack-zun | 14:22 | |
| *** sudipto has quit IRC | 14:41 | |
| *** sudipto has joined #openstack-zun | 14:45 | |
| *** sudipto has quit IRC | 14:55 | |
| *** sudipto has joined #openstack-zun | 14:56 | |
| *** sudipto has quit IRC | 15:03 | |
| *** xiangxinyong has quit IRC | 15:07 | |
| *** janki has joined #openstack-zun | 15:20 | |
| *** sudipto has joined #openstack-zun | 15:41 | |
| *** mkrai has joined #openstack-zun | 15:52 | |
| *** janki has quit IRC | 15:57 | |
| *** sudipto has quit IRC | 16:07 | |
| *** mikelk has quit IRC | 16:12 | |
| *** sudipto has joined #openstack-zun | 16:34 | |
| *** sudipto has quit IRC | 16:42 | |
| *** sudipto has joined #openstack-zun | 16:42 | |
| *** chandankumar has joined #openstack-zun | 16:45 | |
| *** sudipto has quit IRC | 17:28 | |
| *** sudipto has joined #openstack-zun | 17:44 | |
| *** sudipto has quit IRC | 17:44 | |
| *** chandankumar has quit IRC | 18:28 | |
| *** mkrai has quit IRC | 18:47 | |
| *** yuanying has quit IRC | 19:03 | |
| *** flwang1 has quit IRC | 20:11 | |
| *** flwang1 has joined #openstack-zun | 21:32 | |
| *** openstackgerrit has quit IRC | 22:03 | |
| *** openstackgerrit has joined #openstack-zun | 22:04 | |
| *** openstackgerrit has quit IRC | 22:48 | |
| *** openstackgerrit has joined #openstack-zun | 22:48 | |
| *** hongbin has quit IRC | 23:18 | |
| *** shubhams has quit IRC | 23:20 | |
| *** yuanying has joined #openstack-zun | 23:45 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!