| *** yamamoto_ has quit IRC | 00:34 | |
| *** gongysh has quit IRC | 01:24 | |
| *** yamamoto_ has joined #openstack-dragonflow | 01:34 | |
| *** yamamoto_ has quit IRC | 01:43 | |
| *** yamamoto has joined #openstack-dragonflow | 02:07 | |
| *** gongysh has joined #openstack-dragonflow | 02:09 | |
| openstackgerrit | Fei Rao proposed openstack/dragonflow: The implementation of distributed-DNAT https://review.openstack.org/289148 | 02:20 |
|---|---|---|
| *** yamamoto has quit IRC | 02:27 | |
| *** gongysh has quit IRC | 02:41 | |
| *** yamamoto has joined #openstack-dragonflow | 03:09 | |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Implement DB consistency https://review.openstack.org/282290 | 04:21 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Implement DB consistency https://review.openstack.org/282290 | 04:23 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Implement DB consistency https://review.openstack.org/282290 | 06:04 |
| openstackgerrit | hujie proposed openstack/dragonflow: Implement ovsdb monitor function for dragonflow https://review.openstack.org/290353 | 06:27 |
| openstackgerrit | hujie proposed openstack/dragonflow: Implement ovsdb monitor function for dragonflow https://review.openstack.org/290353 | 06:28 |
| openstackgerrit | hujie proposed openstack/dragonflow: Implement ovsdb monitor function for dragonflow https://review.openstack.org/290353 | 06:30 |
| openstackgerrit | hujie proposed openstack/dragonflow: Implement ovsdb monitor function for dragonflow https://review.openstack.org/290353 | 06:30 |
| *** CrayZee has quit IRC | 06:54 | |
| *** CrayZee has joined #openstack-dragonflow | 06:55 | |
| openstackgerrit | Omer Anson proposed openstack/dragonflow: Add DHCP applicative testing https://review.openstack.org/291196 | 06:55 |
| *** yuli_s has joined #openstack-dragonflow | 06:58 | |
| *** oanson has quit IRC | 07:02 | |
| *** gampel has joined #openstack-dragonflow | 07:05 | |
| openstackgerrit | Gal Sagie proposed openstack/dragonflow: Multiple consumers for PLUGIN topic https://review.openstack.org/292743 | 07:13 |
| openstackgerrit | Yuli proposed openstack/dragonflow: Fix file limit bug https://review.openstack.org/292391 | 07:24 |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Increase etcd read timeout https://review.openstack.org/292752 | 07:48 |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Add publish-subscribe abstraction spec https://review.openstack.org/263733 | 07:51 |
| openstackgerrit | Merged openstack/dragonflow: Add Redis_driver.rst https://review.openstack.org/274340 | 08:07 |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Call sync on local cache only on none normal Exception https://review.openstack.org/292447 | 08:28 |
| *** gampel has quit IRC | 08:40 | |
| *** gampel has joined #openstack-dragonflow | 08:41 | |
| *** DuanKebo has joined #openstack-dragonflow | 09:01 | |
| openstackgerrit | hujie proposed openstack/dragonflow: Implement ovsdb monitor function for dragonflow https://review.openstack.org/292768 | 09:05 |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Add publish-subscribe abstraction spec https://review.openstack.org/263733 | 09:07 |
| openstackgerrit | duankebo proposed openstack/dragonflow: Code for selective topology distribution https://review.openstack.org/281687 | 09:22 |
| yuli_s | Hello everyone | 09:27 |
| yuli_s | i am researching the performance degradation issue I found at : https://bugs.launchpad.net/dragonflow/+bug/1556807 | 09:28 |
| openstack | Launchpad bug 1556807 in DragonFlow "Performance degradation when multiple subnets added" [High,New] - Assigned to Yuli (stremovsky) | 09:28 |
| yuli_s | one of the problems I see (it is just one of the issues ) is with the arrays that we store in etcd database objects | 09:30 |
| yuli_s | for example with the lrouter object | 09:31 |
| yuli_s | we have an array of ports | 09:31 |
| yuli_s | and if I look at add_lrouter_port() function, | 09:31 |
| yuli_s | the code fetches object, performs json decode, add new port to the list of ports, stringify back to json object and stores the data back to db | 09:32 |
| yuli_s | when we have a lot of ports like I do have in stress test (i have hundreds of ports in router) | 09:33 |
| yuli_s | this becomes an issue | 09:33 |
| gampel | yuli_s: we do it only on one port so why do you see a difference when there are many ports in the DB | 09:35 |
| gampel | Everyone please review https://review.openstack.org/#/c/281687/19 so we could merge it | 09:35 |
| yuli_s | hm | 09:36 |
| yuli_s | on each new port, the code has to code with more data | 09:36 |
| yuli_s | the json strings are bigger, more memory required to decode them then | 09:37 |
| yuli_s | we encode them back to string | 09:37 |
| yuli_s | i suppose, we might need to conside storing ports for each router as an array | 09:38 |
| yuli_s | currently we have /lrouter/<router-id> with a json | 09:38 |
| yuli_s | we might add /lrouter/<router-id> | 09:39 |
| yuli_s | and /lrouter/<router-id>-ports | 09:39 |
| yuli_s | so, the new version of the add_lrouter_port() might be very simple | 09:41 |
| yuli_s | instead of reading lrouter and decoding, etcc - we will just push new port to /lrouter/<router-id>-ports | 09:42 |
| gsagie | yuli_s: thats a good point | 09:51 |
| gsagie | how many ports are you adding per router? | 09:51 |
| gsagie | how many networks | 09:51 |
| gsagie | as you suggest it might be better to hold them in their own table and not in the router object | 09:52 |
| gsagie | that can explain the behavior you are seeing | 09:53 |
| gampel | yes yuli is now checking the test only with allot of subnet no routers port | 09:54 |
| gsagie | this is the same case | 09:54 |
| gsagie | as we add subnets to network | 09:55 |
| gsagie | With nick-ma work on the lock we can now split these objects | 09:55 |
| gampel | i mean multi networks with one subnet each | 09:56 |
| gsagie | ahh ok | 09:56 |
| yuli_s | i have may be 10 networks | 09:57 |
| yuli_s | with 100 subnets per network | 09:57 |
| yuli_s | each one is added to router | 09:57 |
| yuli_s | gampel: I am running test without add_interface_router() now | 10:19 |
| yuli_s | there is no degradation, | 10:20 |
| yuli_s | in additon, it is much faster | 10:20 |
| yuli_s | i created 400 ports in total | 10:21 |
| yuli_s | time spend to create 100 subnets: 74.4845440388 time spend to create 100 subnets: 68.6519298553 time spend to create 100 subnets: 65.8278388977 time spend to create 100 subnets: 68.8804788589 | 10:21 |
| yuli_s | ops, | 10:21 |
| yuli_s | time spend to create 100 subnets: 74.4845440388 | 10:21 |
| yuli_s | time spend to create +100 subnets: 68.6519298553 | 10:22 |
| yuli_s | time spend to create +100 subnets: 65.8278388977 | 10:22 |
| yuli_s | time spend to create +100 subnets: 68.8804788589 | 10:22 |
| yuli_s | the value in secconds | 10:22 |
| *** oanson has joined #openstack-dragonflow | 10:24 | |
| yuli_s | I coppied my comments to the bug: https://bugs.launchpad.net/dragonflow/+bug/1556807 | 10:33 |
| openstack | Launchpad bug 1556807 in DragonFlow "Performance degradation when multiple subnets added" [High,New] - Assigned to Yuli (stremovsky) | 10:33 |
| openstackgerrit | feipeng proposed openstack/dragonflow: Add Redis DB Driver and Redis PubSub Driver https://review.openstack.org/286028 | 10:53 |
| openstackgerrit | duankebo proposed openstack/dragonflow: Code for selective topology distribution https://review.openstack.org/281687 | 12:35 |
| *** gsagie has quit IRC | 12:56 | |
| *** yamamoto has quit IRC | 13:05 | |
| *** gsagie has joined #openstack-dragonflow | 13:13 | |
| *** dingboopt has quit IRC | 13:27 | |
| *** yamamoto has joined #openstack-dragonflow | 13:49 | |
| *** yamamoto_ has joined #openstack-dragonflow | 13:49 | |
| *** yamamoto has quit IRC | 13:53 | |
| *** Guest41345 has quit IRC | 14:03 | |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Code for selective topology distribution https://review.openstack.org/281687 | 14:20 |
| *** oanson has quit IRC | 14:20 | |
| openstackgerrit | Eran Gampel proposed openstack/dragonflow: Increase etcd read timeout https://review.openstack.org/292752 | 14:51 |
| *** yuli_s has quit IRC | 14:55 | |
| openstackgerrit | duankebo proposed openstack/dragonflow: Code for selective topology distribution https://review.openstack.org/281687 | 15:17 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Add DB Consistency spec https://review.openstack.org/268005 | 16:17 |
| openstackgerrit | Merged openstack/dragonflow: Multiple consumers for PLUGIN topic https://review.openstack.org/292743 | 16:33 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Implement DB consistency https://review.openstack.org/282290 | 16:40 |
| openstackgerrit | Li Ma proposed openstack/dragonflow: Implement DB consistency https://review.openstack.org/282290 | 16:44 |
| *** openstackgerrit has quit IRC | 17:14 | |
| *** openstackgerrit_ is now known as openstackgerrit | 17:14 | |
| *** openstackgerrit has quit IRC | 17:15 | |
| *** openstackgerrit_ has joined #openstack-dragonflow | 17:15 | |
| *** openstackgerrit_ is now known as openstackgerrit | 17:16 | |
| *** openstackgerrit_ has joined #openstack-dragonflow | 17:17 | |
| *** oanson has joined #openstack-dragonflow | 18:54 | |
| openstackgerrit | Merged openstack/dragonflow: Call sync on local cache only on none normal Exception https://review.openstack.org/292447 | 19:10 |
| *** oanson has quit IRC | 20:28 | |
| *** oanson has joined #openstack-dragonflow | 20:34 | |
| *** oanson has quit IRC | 21:20 | |
| *** oanson has joined #openstack-dragonflow | 21:27 | |
| *** oanson has quit IRC | 21:52 | |
| *** yamamoto_ has quit IRC | 23:05 | |
| *** yamamoto_ has joined #openstack-dragonflow | 23:51 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!