| *** zerocoolback has joined #kata-dev | 00:46 | |
| *** zerocoolback has quit IRC | 00:46 | |
| *** BetaXO has joined #kata-dev | 01:32 | |
| *** zerocoolback has joined #kata-dev | 02:01 | |
| *** zerocoolback has quit IRC | 03:10 | |
| *** zerocoolback has joined #kata-dev | 03:11 | |
| *** zerocoolback has quit IRC | 03:11 | |
| *** zerocoolback has joined #kata-dev | 03:11 | |
| *** zerocoolback has quit IRC | 03:12 | |
| *** zerocoolback has joined #kata-dev | 03:12 | |
| *** zerocoolback has quit IRC | 03:13 | |
| *** zerocoolback has joined #kata-dev | 03:13 | |
| *** zerocoolback has joined #kata-dev | 03:14 | |
| *** zerocoolback has quit IRC | 03:14 | |
| *** zerocoolback has joined #kata-dev | 03:15 | |
| *** zerocoolback has joined #kata-dev | 03:15 | |
| *** zerocoolback has quit IRC | 03:16 | |
| *** zerocoolback has joined #kata-dev | 03:16 | |
| *** zerocoolback has quit IRC | 03:17 | |
| *** zerocoolback has joined #kata-dev | 03:17 | |
| *** zerocoolback has quit IRC | 03:17 | |
| *** zerocoolback has joined #kata-dev | 03:18 | |
| *** zerocoolback has joined #kata-dev | 03:19 | |
| *** zerocoolback has joined #kata-dev | 03:19 | |
| *** zerocoolback has joined #kata-dev | 03:20 | |
| *** zerocoolback has joined #kata-dev | 03:36 | |
| *** zerocoolback has quit IRC | 03:41 | |
| *** zerocoolback has joined #kata-dev | 03:53 | |
| *** BetaXO has quit IRC | 04:18 | |
| *** coolsvap has joined #kata-dev | 04:25 | |
| *** zerocoolback has quit IRC | 04:28 | |
| *** BetaXO has joined #kata-dev | 04:28 | |
| *** zerocoolback has joined #kata-dev | 04:28 | |
| *** zerocoolback has quit IRC | 04:34 | |
| *** zerocoolback has joined #kata-dev | 04:50 | |
| *** sjas has joined #kata-dev | 04:53 | |
| *** sjas_ has quit IRC | 04:56 | |
| *** zerocoolback has quit IRC | 05:16 | |
| *** zerocoolback has joined #kata-dev | 05:18 | |
| *** BetaXO has quit IRC | 05:36 | |
| *** marcov has joined #kata-dev | 06:38 | |
| *** jodh has joined #kata-dev | 07:05 | |
| *** davidgiluk has joined #kata-dev | 08:02 | |
| *** gwhaley has joined #kata-dev | 08:05 | |
| *** zerocoolback has quit IRC | 09:46 | |
| *** zerocool_ has joined #kata-dev | 09:52 | |
| *** zerocool_ has quit IRC | 09:56 | |
| *** gwhaley has quit IRC | 11:00 | |
| *** davidgiluk has quit IRC | 11:24 | |
| *** davidgiluk has joined #kata-dev | 11:25 | |
| *** zerocoolback has joined #kata-dev | 11:33 | |
| *** devimc has joined #kata-dev | 11:42 | |
| *** gwhaley has joined #kata-dev | 12:12 | |
| kata-irc-bot | <mvedovati> Jenkins build jobs seems to be stuck | 13:29 |
|---|---|---|
| gwhaley | @mvedovati - I have a feeling the server may have run out of disk space. I pinged an Issue on the ci repo to mnaser and fuentess to see if they can check | 13:31 |
| kata-irc-bot | <mvedovati> gwhaley: ok, thanks | 13:31 |
| *** annabelleB has joined #kata-dev | 13:35 | |
| *** annabelleB has quit IRC | 13:37 | |
| kata-irc-bot | <salvador.fuentes> @graham.whaley @mvedovati taking a look | 13:44 |
| kata-irc-bot | <salvador.fuentes> @mnaser, can you help us check on your side? I am getting this error from the openstack plugin: Global instance cap (10) reached while adding capacity for label: ubuntu-1604 | 13:55 |
| *** annabelleB has joined #kata-dev | 14:13 | |
| *** lpetrut has joined #kata-dev | 14:32 | |
| kata-irc-bot | <graham.whaley> hey @jose.carlos.venegas.m - just a heads up. I dropped a note in the arch call notes for today just to do a check that we are all OK on a plan for hotplug/ballooning. istr we mentioned that would be a good topic to do a check on, but didn't see an entry for it ;) | 14:37 |
| kata-irc-bot | <graham.whaley> its fine if the answer is 'will have more info next week' btw ;) | 14:37 |
| davidgiluk | virtio-mem is certainly something to look at for hotplug/ballooning | 14:41 |
| *** lsm5 has joined #kata-dev | 14:51 | |
| kata-irc-bot | <jose.carlos.venegas.m> I have some slides | 14:52 |
| kata-irc-bot | <graham.whaley> @jose.carlos.venegas.m oh, cool :slightly_smiling_face: | 14:53 |
| kata-irc-bot | <xu> cool | 14:53 |
| kata-irc-bot | <jose.carlos.venegas.m> I would like to have more details | 14:55 |
| *** lpetrut has quit IRC | 14:57 | |
| *** marcov has quit IRC | 14:57 | |
| *** fuentess has joined #kata-dev | 15:00 | |
| *** dklyle has joined #kata-dev | 15:04 | |
| *** marcov has joined #kata-dev | 15:12 | |
| *** coolsvap has quit IRC | 15:24 | |
| *** zerocoolback has quit IRC | 16:35 | |
| *** annabelleB has quit IRC | 16:50 | |
| *** annabelleB has joined #kata-dev | 16:56 | |
| *** gwhaley has quit IRC | 16:59 | |
| *** jodh has quit IRC | 17:02 | |
| kata-irc-bot | <archana.m.shinde> @jose.carlos.venegas.m ping | 17:30 |
| kata-irc-bot | <jose.carlos.venegas.m> @archana.m.shinde pong | 17:32 |
| kata-irc-bot | <archana.m.shinde> just had a question about memory hotplug, you mentioned that balooning can return any amount of memory | 17:33 |
| kata-irc-bot | <archana.m.shinde> does that happen immediately? | 17:33 |
| kata-irc-bot | <archana.m.shinde> @jose.carlos.venegas.m ^ | 17:33 |
| kata-irc-bot | <jose.carlos.venegas.m> @archana.m.shinde yes, resize the balloon will reduce the memory of qemu almost instantly (can take a few seconds but that is async you do the qmp call and then the guest kernel does the job | 17:35 |
| kata-irc-bot | <archana.m.shinde> ok | 17:36 |
| kata-irc-bot | <jose.carlos.venegas.m> so if you stress a VM with 4G of ram - and then you resize the balloon to be 2GB you will get a memory reducion in the qemu process | 17:36 |
| kata-irc-bot | <archana.m.shinde> and what is the impact of not returning the memory slots? | 17:37 |
| kata-irc-bot | <jose.carlos.venegas.m> well today is limited the amount of slots - each slot is a posibility to hot add memory so I we run out of slots and the requested memory at the time we run of slots is not meet we wont be able to add that memory | 17:38 |
| kata-irc-bot | <jose.carlos.venegas.m> if you start a container , then update it to 5G (and hotplug those 5G) - then we reduce 1G and then increase to 3GB not slot is needed. But if we increase to something higher than 5G, lets say 6GB an slot will be needed | 17:40 |
| kata-irc-bot | <jose.carlos.venegas.m> so it depends how the user keeps reducing and increasing memory in the container | 17:41 |
| davidgiluk | hopefully virtio-mem will make this better; https://lwn.net/Articles/755423/ | 17:41 |
| kata-irc-bot | <jose.carlos.venegas.m> @davidgiluk cool you are still here | 17:42 |
| kata-irc-bot | <jose.carlos.venegas.m> yes, we talked about a bit this morning the arch meeting | 17:42 |
| davidgiluk | hot plugging with dimms is very hard - well hot*unplugging* is very hard | 17:43 |
| kata-irc-bot | <jose.carlos.venegas.m> I think is a good option to take a look | 17:44 |
| kata-irc-bot | <jose.carlos.venegas.m> does is virtio-mem is considered stable/(and merged?) in qemu | 17:44 |
| kata-irc-bot | <jose.carlos.venegas.m> I'd like to test it locally | 17:45 |
| davidgiluk | not yet; it's on list for review | 17:45 |
| kata-irc-bot | <jose.carlos.venegas.m> so I think this is a good option for the future | 17:45 |
| davidgiluk | if you ask David Hildenbrand I'm sure he'll be glad to help you get going | 17:45 |
| kata-irc-bot | <jose.carlos.venegas.m> for now balloon I think require a few changes in our code base and help us | 17:46 |
| davidgiluk | balloon is 'delicate' | 17:46 |
| davidgiluk | it also has problems like when you start up, you are using all of memory until you balloon it out - so for example if you start 20 containers then you are using all of their memory until you fix it | 17:46 |
| kata-irc-bot | <jose.carlos.venegas.m> you mean the memory used in the startup of each container ? | 17:48 |
| kata-irc-bot | <jose.carlos.venegas.m> that is not returned back ? | 17:48 |
| davidgiluk | no | 17:48 |
| davidgiluk | guests startup with their balloons deflated, so the guest initially has all of it's memory available | 17:49 |
| davidgiluk | you can inflate the balloon to free up host memory; but at startup of the guest, the guest has it all | 17:49 |
| kata-irc-bot | <jose.carlos.venegas.m> but when we say the has it all * that is in theory - because the pages are given to the guest until it uses | 17:50 |
| kata-irc-bot | <jose.carlos.venegas.m> or am I misunderstanding something ? | 17:51 |
| kata-irc-bot | <jose.carlos.venegas.m> the idea mix of both balloon + hot plug by now - not start the VM with a lot of memory but a few memory | 17:52 |
| kata-irc-bot | <jose.carlos.venegas.m> today is to 2G - then if the user request to increase the memory (e.g docker update) we hotplug - if then is needed to reduce -we restrict/return back by updating the balloon | 17:53 |
| davidgiluk | right, the guest has the memory mapped; so yes in theory you can get away with overallocating | 17:54 |
| davidgiluk | and yes balloon+hotplug solves some of it | 17:54 |
| davidgiluk | yes, doing the dance of hotplug + then ballooning | 17:54 |
| kata-irc-bot | <jose.carlos.venegas.m> so those that appouch sounds good to you while we get an stalbe/merged support for virtio-mem - I still looking forward to test it sounds interesting IMO | 17:55 |
| davidgiluk | 'good' is over stating - I don't really like ballooning - and hotplug is OK but remember you can't really unplug | 17:56 |
| davidgiluk | so it's OK if you need something before virtio-mem | 17:56 |
| davidgiluk | (If you're at KVM forum, David has the last slot talking about virtio-mem) | 17:57 |
| davidgiluk | I'm a couple of slots before and we've been proofreading each others slides :-) | 17:57 |
| kata-irc-bot | <jose.carlos.venegas.m> I am not attending lets but would be interesing see the slides when you have it ready - also probably could be a good topic to get our arch meeting someday | 18:00 |
| kata-irc-bot | <jose.carlos.venegas.m> I'll work to test it - I've will ask to you and David later when I start with that :slightly_smiling_face: | 18:04 |
| davidgiluk | great! | 18:04 |
| *** annabelleB has quit IRC | 18:24 | |
| *** eernst_ has joined #kata-dev | 19:03 | |
| *** davidgiluk has quit IRC | 19:14 | |
| kata-irc-bot | <salvador.fuentes> Hi @mnaser, were you able to take a look at the nested virtualization issues? We are still facing them | 19:50 |
| *** annabelleB has joined #kata-dev | 19:59 | |
| *** annabelleB has quit IRC | 20:19 | |
| *** annabelleB has joined #kata-dev | 20:24 | |
| *** changcheng has quit IRC | 20:24 | |
| *** devimc has quit IRC | 20:46 | |
| *** fuentess has quit IRC | 21:08 | |
| *** annabelleB has quit IRC | 21:37 | |
| *** eernst_ has quit IRC | 22:04 | |
| *** eernst has joined #kata-dev | 22:06 | |
| *** eernst has quit IRC | 22:11 | |
| *** annabelleB has joined #kata-dev | 22:22 | |
| *** annabelleB has quit IRC | 22:52 | |
| *** annabelleB has joined #kata-dev | 22:54 | |
| *** annabelleB has quit IRC | 22:54 | |
| *** zerocoolback has joined #kata-dev | 23:44 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!