| *** ysandeep|PTO is now known as ysandeep | 04:38 | |
| *** ysandeep is now known as ysandeep|afk | 05:18 | |
| *** ysandeep|afk is now known as ysandeep | 08:55 | |
| *** dviroel|out is now known as dviroel|rover | 11:28 | |
| *** ysandeep is now known as ysandeep|afk | 12:22 | |
| admin1 | i got dpdk up .. but i cannot get any instances to launch on it .. | 13:57 |
|---|---|---|
| *** ysandeep|afk is now known as ysandeep | 14:01 | |
| *** dviroel|rover is now known as dviroel|rover|lunch | 14:58 | |
| *** ysandeep is now known as ysandeep|dinner | 15:13 | |
| spatel | admin1 where are you planning to use DPDK ? | 15:22 |
| spatel | if you planning to run general workload on DPDK based compute host then you are not going to gain anything | 15:23 |
| jrosser_ | jamesdenton: you around? i can look at this ironic stuff some more | 15:24 |
| admin1 | spatel, the node will have just 1 vm optimized for latency | 15:28 |
| spatel | ok but what are you looking? what application are you planning to run on VM which required latency? | 15:29 |
| admin1 | some game from a client | 15:29 |
| admin1 | no idea what game is though .. .. i don't get inside the vm login | 15:30 |
| admin1 | i got the dpdk added to br-vlan following yours and a few other articles | 15:30 |
| admin1 | but i can't see to get any vms spawn in that machine | 15:30 |
| admin1 | i disabled all others and enabled just that one | 15:30 |
| spatel | did you configured hugepages on VM? | 15:31 |
| spatel | DPDK is no fun i would say.. it has lots of knob. if they are just running single VM then go without DPDK.. i don't think they will notice any latency issue. (if they are looking for network latency then use SRIOV) | 15:33 |
| jrosser_ | admin1: as far as i understand there really is no point unless the application inside the VM is dpdk aware | 15:34 |
| spatel | jrosser_ 100% agreed | 15:34 |
| spatel | DPDK is myth :) all document saying increate your network throughput but true story is that only with VM running DPDK base app. | 15:35 |
| admin1 | aah | 15:35 |
| admin1 | ok | 15:35 |
| spatel | increase* | 15:35 |
| admin1 | i will ask them | 15:35 |
| spatel | I spent many many months and many lab to understand that line :) | 15:36 |
| spatel | admin1 even i asked same question in berlin conf. to DPDK folks and they agreed and said still you are using virtio to connecting your VM so there is a bottleneck. | 15:37 |
| spatel | Just use SRIOV period | 15:37 |
| spatel | my eyes on this guy and i believe we can take advantage of high speed networking without DPDK shit - https://docs.openvswitch.org/en/latest/intro/install/afxdp/ | 15:39 |
| spatel | * It is aims to have comparable performance to DPDK but cooperate better with existing kernel’s networking stack * | 15:40 |
| spatel | Anyone running windows on openstack ? I have licensing question | 15:42 |
| admin1 | i am :) | 15:43 |
| admin1 | basically licensing is regional .. so you need to first contact the m$ distributor responsible for your area | 15:44 |
| admin1 | next, you ask them for a datacenter license .. which allows you to cram as many vms as possible in the same hardware which you license per box on per set of cpu | 15:44 |
| admin1 | so if you have a big box and only say 1 windows customers, you say you will take min license for 8 cores and then once the 8 core is used(abused), then add 8 core more and grow up | 15:45 |
| admin1 | that is how public cloud providers do it .. | 15:45 |
| admin1 | alternate is to ask customer to bring their own license | 15:45 |
| jamesdenton | hi jrosser_ | 15:50 |
| spatel | great! so if i get datacenter lic for core base in that case how my VM understand HOST is licensed ? | 15:50 |
| jrosser_ | hello | 15:50 |
| jamesdenton | i will be in and out for a bit | 15:51 |
| jrosser_ | ok | 15:51 |
| jrosser_ | i guess the patch with the biggest changes / questions is this one https://review.opendev.org/c/openstack/openstack-ansible-os_ironic/+/852122 | 15:52 |
| jamesdenton | yes | 15:52 |
| jamesdenton | with containers, it seems the bridge IP is not being picked up in inventory. i'm wondering if the dynamic address mechanism is/was meant to work around that in those other two projects | 15:53 |
| jrosser_ | i have this https://paste.opendev.org/show/b3OlRo8NGe0MmgwrmStV/ | 15:55 |
| jrosser_ | and yes we seem to have split universe with dynamic_address_fact and then something similar/different in the octavia role, and now also different in my ironic patch | 15:56 |
| jamesdenton | yes, but conductor actually runs on metal IIRC | 15:56 |
| jamesdenton | and conductor is the one that matters here for tftp/http | 15:56 |
| jamesdenton | that is to say, even w/ lxc i think conductor still runs on metal | 15:57 |
| jrosser_ | https://paste.opendev.org/show/bu5VE8Bj5FioJK3o3c8g/ | 15:58 |
| jrosser_ | do you have an AIO for this so we can compare? | 15:58 |
| jamesdenton | i think so, one sec | 15:59 |
| jrosser_ | also https://paste.opendev.org/show/bUCstZrdgCQBSlV3MNz2/ | 16:01 |
| jamesdenton | i'm rolling with this: https://github.com/openstack/openstack-ansible/blob/master/inventory/env.d/ironic.yml#L31-L39 | 16:02 |
| jrosser_ | hmm interesting | 16:04 |
| jamesdenton | having to reboot this VM | 16:04 |
| jamesdenton | jrosser_ https://paste.opendev.org/show/bsmNoGFtOYd4S5WdXof6/ | 16:09 |
| jamesdenton | the behavior of bmaas_address in this case is just like storage and tunnel_address | 16:10 |
| jamesdenton | this environment is a little screwy, since I modified is_metal from true to false to get it to build an ironic_api container. otherwise, there wasn't one | 16:11 |
| jrosser_ | how wierd | 16:11 |
| jamesdenton | and there is a patch out there.. it was intentional | 16:11 |
| jamesdenton | https://github.com/openstack/openstack-ansible/commit/4cba47c1d3a186d189c75e903cbb8be39602fa0f | 16:11 |
| *** dviroel|rover|lunch is now known as dviroel|rover | 16:12 | |
| jamesdenton | well, it's quasi-related, anyway. | 16:13 |
| jrosser_ | i'm just wondering how i have ended up with ironic_api container at all here | 16:14 |
| jamesdenton | good question | 16:14 |
| jrosser_ | maybe another good question is what actually do we want | 16:20 |
| jrosser_ | my AIO is from before i found the weird inventory bug, so it could be that i also have a non representative state | 16:21 |
| *** ysandeep|dinner is now known as ysandeep|out | 16:57 | |
| damiandabrowski | looks like we have a problem with rabbit repo, at least on 24.4.0 + focal | 17:13 |
| damiandabrowski | https://paste.openstack.org/show/biMEpiqFoTfuicwDLilS/ | 17:13 |
| damiandabrowski | 1:24.1.3-1 disappeared from the repo? | 17:14 |
| damiandabrowski | https://paste.openstack.org/show/byAFbw5rwUsjF8ZjpQLB/ | 17:14 |
| damiandabrowski | seems like it's available only for ubuntu/bionic and debian/buster | 17:17 |
| damiandabrowski | https://cloudsmith.io/~rabbitmq/repos/rabbitmq-erlang/packages/?q=name%3A%27%5Eerlang-base%24%27+version%3A1%3A24.1.3-1&sort=-version | 17:17 |
| damiandabrowski | will continue tomorrow | 17:17 |
| *** dviroel|rover is now known as dviroel|out | 21:53 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!